1?

firm“... .

3.
11¢
3...}. .
“antenna?
...IID‘I .»

03.. arr...

If»?

at ...y
..l q .
tfgititi

2l¢
-L’uur.
:5
3
«Jr
’ig

3'38...
......II:
. . . > CI
2.3. . .
... .1. . .

d. 1 . .
...:zz .. new...» {I «3.2)
..t'nkli. . 13.3.49. :1 “a. .3
.. 5...? $3.43... ...dmwlif,
.vvax a: ..3 ’13:...—

. . ...... z; .-.

...“:-
z... E. .. $1
1.1.0.2.. .hr

\ . L .1 3 :1
u ....th H3. riltmaul... .. w.
a: . 3.!
14:123.; $5}... .3.“
5 . #39:! 3. .90 1} J1
‘Llevpr’lLyakmxyangtui 3
...l... ... ...{llsrxau;§
.15. Mafiafifi: I: 1.1.3:; -. .— 5
I3: : . 32a .51.; 5.1.1
Sfivfi‘diafifl in}... a 30‘ 039...“, ...
3. .... ... . t?
5:... :3.th i 7 : .‘1‘uo«|lzav»fr
5.11 . 3-3)}... bats-...!
. . 7 {31‘}:

......fio a.
Is...” .....Sh Advh‘:il.§
. raft. .st!:r..$u pittii’x
!’ . '1‘! ’V.§2Jfi( ’5
nag»... . .. 513...»???
I. .. , . «(heaveacfi
41v

 

». 1’9.
1....” l... 19:: J
‘ {...
2...... .

.... . I- I: n .3}.
Au‘ d.£$)

:1 3 A. .
1,! {thiginf : ‘1
.... :53...
....EA: 7‘!»
End ..a

Qatlv 3131.3.
3:33.! 131:. v’.
‘Ibafi

>2. .03.. .
1:: 3...) 3,
125.;

“20.64:

 

‘5:
. . . u "this.
. . . . >. ...,mar ludhd. ul
. . :13. Z» 3!. . :153; i
. . . . z L...~..¢.ollfi‘1§.£t¢.f\4(
. . . ,u :3... r}...
x. h. . a»... . . ism...
. .34.»... . . . .
ma}

 

......H..:..£. ..t a 5?. .

THESIS
Q
20 O I

51/00 €702? LIBRARY
Michigan State
University

 

 

 

 

This is to certify that the

dissertation entitled
RISK MARKERS 0r INTIMATE ASSAULTERS 0N DANGER ASSESSMENT (DA)

INSTRUMENT

--- TOWARD A CROSS-CULTURAL UNDERSTANDING FROM TAIWAN SAMPLES

presented by
Min-chieh Lin
has been accepted towards fulfillment

of the requirements for

PhD. degree in Criminal Justice

 

 

%/

Jpréfessor

 

flq 2.00 / Charles J, (‘orlcy
Date 8},

MS U i: an Affirmative Action/Equal Opportunity Institution 0- 12771

 

PLACE IN RETURN Box to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECAUID with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRCIDatoDuepes-sz

RISK MARKERS OF INTIMATE ASSAULTERS ON DANGER ASSESSMENT
(DA) INSTRUMENT
TOWARD A CROSS-CULTURAL UNDERSTANDING FROM TAIWAN
SAMPLES
By

Min-chieh Lin

AN ABSTRACT OF A DISSERTATION
Submitted to
Michigan State University
In partial fulfillment of the requirements
For the degree of
DOCTOR OF PHILOSOPHY
School of Criminal Justice

2001

Associate Professor Charles J. Corley

ABSTRACT

RISK MARKERS OF INTIMATE ASSAULTERS ON DANGER ASSESSMENT (DA)
INSTRUMENT
TOWARD A CROSS-CULTURAL UNDERSTANDING FROM TAIWAN
SAMPLES
By

Min-chieh Lin

Taiwan passed the Domestic Violence Prevention Law in 1998 and implemented it
in 1999. The main purpose of this law is to prevent intimate abuse. Taiwan has not
designed a risk assessment tool yet. This research used the most famous risk assessment
tool in the US, the Danger Assessment (DA). It was completed by female victims in
Taiwan, to assess the severity risk markers of intimate abuse among 3 Taiwan sample.
The physical violence part of the Conflict Tactics Scale (Brief CTS) was used to assess
the amount of intimate violence behaviors. For ethical reasons, this study did not contact
the abusers. The author sampled the reporting and willing victims in Domestic Violence
Centers in two local areas in Taiwan during a 4-week period, and requested the social
workers in the two Centers help victims to fill out the DA, the Brief CTS, and the Victim
Questionnaire. Multiple regression analysis was used to test whether all of the items in

the DA and the other risk markers in the Victim Questionnaire to identify predictors of

the severity of intimate assault in Taiwan. Brief CTS scores were used as dependent
variables. It was found that the DA scores in this study were in a normal distribution,
whereas the Brief CTS scores were in poisson distribution. By using poisson regression,
it was found the DA could explain 29.26% of total variance in the amount of violence in
the Brief CTS. In the predictive accuracy test for predicting the appearance of lethal
violence, the area under the ROC curve of the DA was .753, which was a moderate and
satisfactory accuracy, and the area under the ROC curve of the Brief CTS was .718,
which was a lower moderate and still satisfactory accuracy. A cluster analysis found that
the 4 clusters of intimate assaulters could be identified in Taiwan sample. They were
termed: (l) “low-violent assaulter”, (2) “medium-violent, and low-controlling assaulters”,
(3) “high-violent, high-lethal, and alcoholic assaulters”, and (4) “high-violence, high
controlling assaulters”. Based on the findings of this study, it was proposed that
“pathological behavior appearance theory” would help better understand the behavior

patterns of different types of male intimate assaulters in Taiwan.

Copyright by
Min-chieh Lin
2001

In Memory of My Parents

AKNOWLEDGEMENTS

I am extremely thankfiil to my all committee members of my dissertation, Dr.
Charles Corley, Dr. Vincent Hoffman, Dr. Andrew Barclay, and Dr. Alfiee Breland. Dr.
Corley gave me the wonderful opportunity to be his teaching assistant for teaching
statistical computer software, and also helped me to go through some difficult time
during this study. Dr. Hoffman always gave me some wise words whenever I needed. Dr.
Barclay, a clinical psychology professor, assisted me to learn how offenders going on in
their psychopathological problems. Dr. Breland, a counseling professor and also my
adviser in counseling master program, help me refine my counseling skill and knowledge
a lot during the years I studied at MSU.

I also thank Dr. Donald Wilde, the director at Total Health Education Inc., a private
and model counseling center for probationers and parolees, in northern Lansing,
Michigan. He gave me the wonderful opportunity to be an intern in his counseling center
and helped me to learn the community treatment programs for sex offenders, domestic
violence abusers and, substance abusers, all of which are invaluable experience for me

and even for my Taiwan government. Ms. Renee Holsteirn LPC, my counseling intern

vi

supervisor, always taught me the counseling skill without reservation. I really thank her.

Mr. Jeng-daw Tsai, the executive secretary of Taiwan Domestic Violence Council
Ministry of Interior, and Ms. Shin-yuan Chang, the section chief in the Council, are so
wonderful to grant me the official consent for conducting this study in Taiwan. I want to
send my highest regard to them!

All social workers in the Domestic Violence Prevention Centers in Kaohsiung City
Government and Kaohsiung County Government assisted me to collect the data from the
willing battered women, even thought they were really busy. Without their kindest help, I
really could not finish this study. Besides, because of their efforts on their jobs, they are
making the homes safer for the battered women. They all deserve my greatest gratitude
and respect. Special thanks also for Hui-ying Liu and Mei-shu Liu, the director of
Kaohsiung City and County Centers, Hsiu—chin Tan, Li-fun Ten, and Chi-ming Deng, the
area supervisors in Kaohsiung County, and Mei-o Chuang, the senior social worker in
Kaohsiung City, since they were so kind to assist me collect the data. I also want to thank
Chih-chung Huang, a medical doctor focused on intimate violence in Kaohsiung, since he
gave me precious suggestions for my study.

Chien-Ho Wang, and Mao-shen Chen, the Ph.D. candidates in economics at MSU

and my Taiwan friends as well, introduced me some concepts of econometrics and

vii

decision theory that help me to move smoothly on this study. Chien-ming Tseng, the Ph.D.
student in educational statistics department at MSU, gave me some statistical suggestions.
Chia—su Hsu, Pei-Chen Hsu, Ya-chien Wang, and Chien-ming Tseng assisted me to
evaluate the face validity and content validity of the instruments in this study. They are so
wonderful, and deserve my highest appreciation.

My greatest thank also really needs to go for my oldest brother, Min-cheng Lin. He
helped me financially to go through the study at MSU, and said that he helped me
because he had to substitute my pass-away father’s job, to support me. I should not let
him down, and probably I did not with earning a' Ph.D. degree for the Lin’s.

Finally I think I really want to thank all of the above people again, and said I will

really miss all of you in my whole life!

viii

TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................... xi
LIST OF FIQURES ......................................................................................................... xiii
CHAPTER 1
INTRODUCTION ....................................................................................... , ..................... 1
Statement of the problem .......................................................................................... 1
Purpose of this study ................................................................................................. 3
Definition of the terms .............................................................................................. 3
Study Questions in this Research .............................................................................. 4
CHAPTER 2
LITERATURE REVIEW ................. . ................................................................................. 6
The prevalence of intimate violence in the US. and Taiwan ..................................... 6
Risk markers of intimate assault in previous studies ........................... ...................... 7
Introduction of the DA and the CTS .......................................................................... 11
Studies related to the DA ........................................................................................... 19
Issues in measurement of intimate violence .............................................................. 2]
Reliability and validity of the CTS ............................................................................ 23
Causes of intimate assault and related theories ......................................................... 25
Diverse ethnic group in Taiwan ................................................................................. 29
Diverse cultures of marriage in Taiwan ..................................................................... 31
Diverse Religions in Taiwan ...................................................................................... 35
Cultural differences between the US. and Taiwan in literature ................................ 36
Development of risk assessment of violent offenders ............................................... 39
Typology of male intimate assaulters ........................................................................ 41
CHAPTER 3
METHOD .......................................................................................................................... 44
Hypotheses .................................................. . .............................................................. 44
Measurement tools ..................................................................................................... 45
Sampling process and data collecting method ........................................................... 48

ix

Translation procedures and reliability test of research instruments .......................... 50

CHAPTER 4

RESUTLS .......................................................................................................................... 55
Demographic analyses ............................................................................................... 55
General finding in DA and Brief CTS Scores ....................................................... . ..... 6 0
Differences of 15 DA Items between assaulters with and without lethal violence....64
Predicting Amount of Violence in Brief CTS by DA (Poisson regression) ............... 65
Lethal Violence and Its Analyses ............................................................................... 71
Six Other Important Variables in Domestic Violence ................................................ 89
Whether Violent Outside Could Be a Criterion to Distinguish the Different Types of
Batterers ..................................................................................................................... 97
Cluster Analysis of Intimate Abusers by the 15 DA Items ....................................... 99
Factor Analysis of Abusers’ Pathological Behaviors in DA .................................... 104
Civil Protection Order .............................................................................................. 105

CHAPTER 5

DISCUSSION .................................................................................................................. 1 1 1
Summary ................................................................................................................. 111
Toward an Understanding of the Risk Assessment in Taiwan Male Intimate

Assaulters: A Grounded Theory ........................................................................... 114

Treatment Implication for Different Types of Intimate Assaulters .......................... 115
Advantages of this study .......................................................................................... 119
Disadvantages of this study ..................................................................................... 119
Suggestions for future study .................................................................................... 120

APPENDICES ................................................................................................................. 123

BIBLIOGRAPHY ............................................................................................................ 13 l

LIST OF TABLES

Table 1 The Prevalence of Intimate Violence in the US. and Taiwan ............................. 7
Table 2 Danger Experienced by Attempted F emicide and Partner Abuse Controls by
DA ................................................................................................................................ 14
Table 3 Reliability and Validity in Danger Assessment (DA) Research ......................... 21
Table 4 Alpha Reliability Coefficients for CTS .............................................................. 23
Table 5 Correlation of Spousal Report CTS Scores with Student Report CTS Scores..24
Table 6 Reliability Test of Chinese Version Danger Assessment ................................... 52
Table 7 Reliability Test of Chinese Version Brief Conflict Tactics Scale ....................... 53
Table 8 Populations, Numbers of Monthly/Yearly Cases, and Numbers of Samples in
Kaohsiung City and Kaohsiung County ....................................................................... 55
Table 9 Demographic Descriptions of Female Victim Respondents in This Study ........ 58
Table 10 Demographic Descriptions of Female Abuser Respondents in This Study ..... 59
Table 11 Frequency and Percentage in 9 Brief CTS Items ............................................. 62
Table 12 Amount of Violence and its Relationship to Demographic Data. .................... 64
Table 13 Mean Differences between Abusers with and without Lethal Violence in 15
DA Items among Taiwan Sample by T-test ................................................................. 65
Table 14 Predicting the Amount of Violence (Brief CTS) by 15 Items in Danger
Assessment by Poisson Regression .............................................................................. 69
Table 15 Forward Stepwise Regression Analyses on 15 DA Items to Predict the
Amount of Violence in Brief CTS ............................................................................... 69
Table 16 Predicting the Amount of Violence (Brief CTS) by 12 Items in Danger
Assessment scale by Poisson Regression (excluded Item 5, 8, and 10) ..................... 70
Table 17 Five Principle Components among the 12 Items in Danger Assessment
scale .............................................................................................................................. 70
Table 18 Predicting the Amount of Violence (Brief CTS) by 5 principle components
among the 12 Items in Danger Assessment scale by Poisson Regression ................... 71
Table 19 Area * Lethal violence or not Crosstabulation ................................................. 72
Table 20 Time of Lethal Violence and its Relationship to Demographic Data ............... 73
Table 21 Mean Differences of Brief CTS and DA between Abusers with and without
lethal violence .............................................................................................................. 74
Table 22 Area Under the ROC Curves for Predicting Lethal Violence by CTS and
DA ................................................................................................................................ 77
Table 23 Bivariate Correlation within CTS, DA, Lethal Violence ................................. 77
Table 24 Three Categories for Stratifying the CTS Scores ............................................. 78

xi

Table 25 Descriptions in Scores of Brief CTS and DA .................................................. 78
Table 26 Areas under ROC Curves for three Categories in CTS Scores to Predict the

Appearance of Lethal Violence .................................................................................... 79
Table 27 Three Categories for Stratifying CTS Scores ................................................... 80
Table 28 Areas under ROC Curves for three Categories in DA Scores .......................... 81
Table 29 Pearson R of Bivariate Correlations among 15 DA Item ................................. 84
Table 30 Predicting the Appearance of Lethal Wolence by 15 Items in Danger

Assessment by Logistic Regression ............................................................................. 85
Table 31 Predicting the Appearance of Lethal Wolence by 4 Items in Danger

Assessment by Logistic Regression ............................................................................. 85
Table 32 Area under ROC Curves for Predicting Lethal Violence by Item 1, 3, 6(1), & 12

in DA ............................................................................................................................ 86
Table 33 Predicting the Times of Lethal Violence by 15 Items in Danger Assessment by

Poisson Regression ...................................................................................................... 88
Table 34 The Victims’ Responses on Whether Been Stalked or Not .............................. 91

Table 35 Whether or Not the Abusers Threatened to Hurt Victims’ Family of Origin..93
Table 36 Whether the Victims Lived with their Mother-in-laws in Urban and Rural

Areas ............................................................................................................................ 95
Table 37 Whether Vlctims’ Living and Arguing with Mother-in-laws Correlate the
Amount of Violence and Appearance of Lethal Violence ........................................... 96
Table 38 Six Other Variables’ Correlation and Prediction on the Amount of Violence
and Appearance of Lethal Violence ............................................................................. 96
Table 39 Forward Stepwise Poisson Regression Analyses on the 12 DA Items to Predict
the Amount of Violence in Brief CTS ......................................................................... 97

Table 40 Correlation Between Whether Violent Outside and Other DA 14 Items ......... 98
Table 41 Comparing Means on Brief CTS, DA, and Times of Lethal Violence between

the Abusers with and without Outside Violence ................................................. . ........ 99
Table 42 Classification Results in Stepwise Discriminant Analysis ............................. 102
Table 43 Difference of Variables Among the 4 Clusters on CTS, Lethal Violence, DA,

15 DA Items, and Other Variables ............................................................................ 103
Table 44 Factor Analysis of Abusers’ Pathological Behaviors 1n DA Items related to

Abuser’ s Behaviors (excluding Item 6(2) and 11) .................................................... 105
Table 45 Whether the Victims Have Protection Order .................................................. 106
Table 46 ANOVA on CTS, DA, and Times of Lethal V101ence among Victims’ different

perception on effectiveness of protection order ......................................................... 108

Table 47 Victims’ Worries about the Protection Order in Multiple Response Scale....109

xii

LIST OF FIGURES

Figure 1 Map of Taiwan ................................................................................................ 30
Figure 2 Receiver Operative Characteristic (ROC) ...................................................... 75
Figure 3 Predicting Lethal Violence by DA and CTS in ROC Curve Prediction
Accuracy ..................................................................................................................... 76
Figure 4 Predicting Lethal Violence by three Categories of CTS in ROC Curve
Prediction Accuracy ..................................................................................................... 79
Figure 5 Predicting Lethal Wolence by three Categories of DA Scores in ROC Curve
Prediction Accuracy ..................................................................................................... 81
Figure 6 ROC Curves for Predicting Lethal Violence by Item 1, 3, 6(1), & 12 in DA..86
Figure 7 Histogram of Times of Lethal Violence in Taiwan Samples ............................ 88

xiii

CHAPTER 1 INTRODUCTION

StatementhftheProblem

According to the Women Life Situation Survey at Taiwan Province in 1992, about
17.8% of married women had been physically abused in Taiwan. Among them, about
0.2% responded that they could not tolerate it any more, whereas 16.4% responded it
happened once in a while, and 1.2% responded it happened frequently (Department of
Social Afi‘air, Taiwan Province Government, 1991). This did reveal a very serious
phenomenon among Taiwan married women. The spousal assaults identified by
American studies show that abuse could seriously hurt both the physical and
psychological health of women (Walker, 1984; Campbell, 1995), and even influence the
mental health of the children raised in these abusive environments (Hotaling & Sugarman,
1986). In 1993, a Taiwanese woman who was repeatedly battered, Ms. Deng, killed her
husband while he was asleep. Ms Deng was sentenced to 3 years in prison for the murder
(Chou, 1995). However, this case caused heated discussions in Taiwan, and finally caused
the Legislative Yuan (the Taiwan Congress) to pass the Domestic Violence Prevention

Law indirectly in 1998. The main purpose of this law is to prevent intimate abuse, even

though it defines domestic violence as violence between any family members. Taiwan
just began implementing this law as late as 1999, and there are many related prevention
programs still being designed, such as the abusers’ treatment program and risk assessment
tools. Already established are civil protection order procedures, the abused women’s 24-
hour hotline, and local domestic violence prevention centers.

Taiwan has not designed a risk assessment tool yet. It may be appropriate to
introduce some well-developed foreign risk assessment tools to Taiwan, reassess their
validity specifically among the Taiwan sample, and then adjust them for Taiwan use.
However, using foreign risk assessment tools would present some challenges because of
the cultural differences between the culture of origin for these tools and the Taiwanese
culture.

Anthropological and cross-cultural studies have shown that even though spousal
assault is common around the world, there are still variations in its definition, acceptance,
frequency, characteristics, and causes across different societies (Campbell, 1999).
Therefore, exploring the cultural difference of the risk markers of intimate assault
between America and Taiwan would be very important, especially if Taiwan wants to

introduce the American risk assessment scale in Taiwan.

W

The purpose of this study was to see if the Danger Assessment (DA) scale can be
used to accurately predict the lethality of intimate assault cases in Taiwan. This study
would be helpful in developing a risk assessment tool for Taiwan.

According to Campbell, there are two different types of risk assessment for
intimate violence, recidivism risk assessment and lethality risk assessment. Recidivism
risk assessment is the risk assessment for predicting the future recidivism rate, and
lethality risk assessment is the risk assessment for predicting the severity of future assault
or the so-called lethality (personal communication, March 24, 2000). Originally, the
Danger Assessment scale was designed as the risk assessment for lethality in intimate
assault. Moreover, the DA can be also used as the risk assessment for predicting future
recidivism (Goodman, Dutton, & Bennett, 2000).

Assessing lethality of intimate assault is an extremely important issue, because the
social service agencies and the police department need to intervene in the intimate assault
cases and assure that the female victims’ lives are protected. Therefore, this study would

only focus on assessing the lethality risk for intimate assaulters in Taiwan.

Intimate assaulters in this study was defined as a male using any intentional act or
series of acts that was or were physical violence toward his intimate partner.

Risk markers of intimate abuse are the correlational factors associated with the risk
of men who act violently against their female intimate partners (Aldarondo & Sugarman,
1996). It is noted here that the risk markers may not be the causal factors of intimate
abuse.

Lethality risk in this study was defined as the risk that the male intimate assaulters

used lethal violence toward their female intimate partners.

As mentioned earlier, assessing the lethality of intimate assault is an extremely
important issue while the government intervenes in intimate assault cases and protects the
victims’ lives. This study would focus on developing a lethality risk assessment, which
could be used in Taiwan in the future. This research would try to use the Danger
Assessment (DA), the most famous risk assessment tool in America, in Taiwan areas and
assess its performance for the Taiwan sample. Therefore, the study questions of this
research would be: how accurate the DA is in assessing the lethality of Taiwan intimate

assault cases and are there any other risk markers based on the cultural difference that can

be found in this research. Moreover, since it was previously found that inmate assaulters
are not a homogeneous group and typologies have been developed (Holtzworth-Munroe
& Stuart, 1994), it would be worthwhile to learn whether the different types of Taiwan

intimate assaulters as well as their lethality risk could be identified in this study

CHAPTER 2 LITERATURE REVIEW

MS. According to two landmark studies regarding family violence in the
US, done by the Family Research Laboratory at the University of New Hampshire, the
National Family Violence Survey of 1975 (using CTS Form A) and the National Family
Violence Resurvey of 1985 (using CTS Form R), it was found that for the husband-to-
wife violence in 1975 general violence happened among 121 couples and severe violence
happen among 38 couples per 1000 couples, whereas in 1985 general violence happened
among 113 couples and severe violence happened among 30 couples per 1000 couples
(Straus & Gelles, 1986). The general violence in the CTS included threw, push, grab,
shove, and slap, whereas the severe violence included kick, bit, hit, with fist or something,
beat-up, choke, threatening with a knife or gun, and using knife or firing a gun.

Taiwan. Two official surveys for women’s life were done by the Social Affair
Department, Taiwan Province Government (1990, 1992) (not including Taipei City and
Kaohsiung City) in 1990 and 1992, and the sample sizes were 2000 and 1800,

respectively. It was found that in 1990 when married women were asked whether been

abused or not, 11.5% responded “once in a while”, 0.9% responded “usually”, and 0.1%

responded “it was intolerable”. Overall 12.5% of married women responded that they

were abused by their husbands. The 1992 Survey found that 16.4%, 1.2%, and 0.2%,

respectively, and overall, 17.8% of married women responded that they were abused by

their husbands. It is noted that these two surveys in Taiwan did not use the CTS, so it

could not be compared between Taiwan and the US. (Table 1).

Table 1 The Prevalence of Intimate Violence in the US. and Taiwan

 

 

 

 

 

U.S. Taiwan
1975 1985 1990 1992
N= 2143 N=3520 N= 2000 N= 1800
0 General violence 0 General violence 0 Once in a while 0 Once in a while
12.1% 11.3% 11.5% 16.4%
0 Severe violence 0 Severe violence 0 Usually 0 Usually
0.38% 3.0% 0. 9% 0. 2%
0 Almost intolerable 0 Almost intolerable
0.1% 0.2%
Total 15.9% Total 14.3% Total 12.5% Total 17.8%

 

Note. The US. data was from National Family Violence Survey of 1975 and National
Family Violence Resurvey of 1985 (see Straus & Gelles, 1986). Taiwan data was from
Taiwan Province Government (1990, 1992)

3.1“] El' 5 l'E' 51'

By sampling the battered women with and without homicidal behavior toward their
male partners (sample sizes were 42 and 205, respectively), Browne (1984) concluded

that an apparent relationship does not exist between lethality and a number of factors

common in battering relationship. When statistically comparing domestic violence

homicide and abuse-only cases, Browne found a number of factors that distinguished

these two subject populations: (1) frequency of violent incidents; (2) severity of injuries;

(3) man’s threats to kill; (4) woman’s suicide threats; (5) man’s drug use; (6) man’s

frequency of intoxication; (7) forced/threatened sexual acts.

Hotaling and Sugarman (1986) evaluated 97 potential risk marker of husband to
wife violence by reviewing 52 case-comparison studies. They divided all of the risk
markers into four categories: consistent risk (over 70% of studies support it), inconsistent
risk (31% to 69% of studies support-it), consistent non-risk (less than 30% of studies
support it), and risk markers with insufficient data support (less than 3 studies support it).
The risk markers were also classified into three dimensions: wife characteristics, husband
characteristics, and couple characteristics. It was found as the following.

(1) Wife characteristics: (a) consistent risk marker: Only one was found——witnessing
violence as a child or adolescent (73% of studies); (b) inconsistent risk marker:
experiencing violence as a child or adolescent (69% of studies), drug abuse (60% of
studies), low self—esteem (60%), low education level (54%), traditional sex role
expectation (50%), young age (42%), race (3 8%), low assertiveness (33%), low

personality integration (33%). (0) consistent non-risk markers: housewife status

(25%), alcohol usage (17%), and income (0%).

(2) Husband characteristics: (a) consistent risk markers: There were eight identified. They
were sexually aggressive toward wife/partner (100%), violence toward their children
(100%), witnessing violence as a child or adolescent (88%), occupational status
(80%), alcohol usage (78%), low income (75%), low assertiveness (75%), and low
educational level (73%); (b) inconsistent risk markers: experiencing violence as a
child or adolescent (69%), unemployment (66%), criminal record (67%), self-esteem
(60%), age (44%), and need for power (33%); (c) consistent non-risk marker:
traditional sex-role expectation (25%).

(3) Couple characteristics: (a) consistent risk markers: There were five been identified.
They were verbal argument frequency (100%), religion incompatibility (100%), low
marital satisfaction (83%), marital status (83%), and family income/social class (78%);
(b) inconsistent risk markers: educational incompatibility (67%), occupational
incompatibility (67%), and length of relationship (40%); (c) consistent non-risk
marker: decision-making power (25%) and number of children (14%).

Aldarondo and Sugarman (1996) evaluated the utility of specific risk markers of
wife assault in understanding the cessation and persistence of violence against women

over a 3-year period. Longitudinal data were used to identify violent men who ceased the

violence for 2 years and violent men who persisted in using violence, whereas a group of
nonviolent men was used for comparison. Factor analysis indicate that marital conflict,
socioeconomic status (SES), and witnessing violence in the family of origin were the
most powerful discriminators of these groups. It was also found that high levels of marital
conflict and low SES were associated not only with the occurrence of violence but also
with the continuity of wife assault over time.

McFarlane, Parke r, and Soeken (1995) did a stratified and prospective cohort study
in urban public health prenatal clinics to determine the frequency, severity, and
perpetrator of abuse during pregnancy as well as the occurrence of risk factors of
homicide. All women subjects were administered the CTS, the DA, and the Index of
Spouse Assault. The result showed the prevalence of physical abuse during pregnancy
was 16%, almost one in every six women. In addition, they found women abused during
pregnancy had significantly higher scores on all instruments and more risk factors of
homicide when compared with women abused prior to but not during pregnancy.
Therefore, they contended that abusing women during pregnancy is a significant risk
marker for the amount of violence as well as the lethality.

In McFarlane, Campbell, Wilt, Sachs, Ulrich, and Xu (1999), one hundred forty-

one femicide and 65 attempted femicide incidents were evaluated. They found a

10

statistically significant association existed between intimate partner physical assault and
stalking of femicide victims as well as attempted femicide victims. Stalking is revealed to
be correlated to lethal and near lethal violence against women and, coupled with physical
assault, is significantly associated with murder and attempted murder. McFarlane et al
highly suggested that stalking must be considered a risk marker for both femicide and
attempted femicide.

Weisz, Tolman, and Saunders’ (2000) findings were from a secondary data analysis
comparing the accuracy of 177 battered women’s predictions of whether they would be
assaulted again to risk factors supported by previous studies. These battered women’s
predictions were associated with recurrence of severe violence in a bivariate analysis.
These predictions also added significantly to the accuracy of established risk factors in
two multivariate equations predicting severe reassault within a 4-month period. The
results also showed that even though not all of the battered women made accurate
predictions, this study supported that the use of battered women’s predictions about the

possibility of being re-abused in the future should be included in risk prediction.

launductionnfltheDAandtheflIS
W). The DA was developed by Jacquelyn Campbell,

11

a resident nurse and a nursing professor at the John Hopkins University. The DA was
originally deve10ped in 1984, and revised in 1995. There are 15 items that have to be
responded from the victims. They are increased frequency of physical violence, increased
severity of physical violence, occurrence of choking, presence of firearms, occurrence of
forced intercourse, drugs usage, threatening to kill victim or victim’s perception of
capability to kill, drinking, controlling daily behavior of victims, physical assault while
pregnant, violent jealousy, suicidal behavior or attempt of both parties, violence toward
children, and violence outside the house (see Appendix 1). Campbell (1995) mentioned
that in five studies the means of the DA have ranged from 5.5 to 8.7 in the sample of
abused women (the presence of each item would be assigned a score of 1, and the
absence would be 0.). Even thought the DA did not provide the formal cutoff point to
decide whether the batterer is dangerous and may commit a homicide, Campbell
mentioned that ten would be a good cutoff point for this purpose (Campbell, 2000). The
reliability and validity of the DA would be discussed in the latter session.

Previous research on the 15 items of the DA were organized as the following.
Regarding increased frequency of battering, men who have demonstrated assaultive
behavior in either past or current intimate relationships are at risk for future violence

(Fagan, Steward, & Hansen, 1983; Sonkin, 1987). It is also found that a pattern of recent

12

escalation in the frequency or severity of intimate assault is associated with imminent risk
for violent recidivism (Hart, 1992; Sonkin,1987; and Stuart & Campbell, 1989).

Regarding increased severity of battering, it is found that a pattern of recent
escalation of in the frequency or severity of intimate assault is associated with imminent
risk for violent recidivism (Hart, 1992; Sonkin,]987; and Stuart & Campbell, 1989).

Presence of choking behavior was not be included in the DA until 1995. Choking is
a behavior that obviously can result in victims’ death. According to Campbell (2000),
when compared attempted femicide and non—femicide intimate assault cases, it was found
that the two groups did reach significant difference in this item (see Table 2).

Guns are the weapons used in the majority (64%) of homicides between family
members in America (Rosenberg & Mercy, 1985). Their easy accessibility in this country
has been linked with the high rates of homicide in the US. while compared with other
countries. Rosenberg and Mercy (1985) further suggested that limiting the availability
and access of guns to the population-at-large or certain high risk subsets is most likely to
reduce killings among spouses and young men. However, as the author would mention in
a cross-culture comparison session, since Taiwan is a firearm-control country and very
few people can be allowed to own firearms, it may not be appropriate to use this item as a

risk marker.

13

Table 2 Danger Experienced by Attempted Femicide and Partner Abuse Controls by DA

 

Attempted Control

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1. physical violence increased in frequency 56%* 24%
2.plysical violence increased in severity 62%* 18%
3. partner tried to choke victim 50%* 10%
4. a gun is presence in the house 34%* 16%
ipartner forced victim to have sex 39%* 12%
6. partner used street drugs 55%* 23%
7(1) Partner threatened to kill victim 57%* 14%
7(2). Victim believes partner is capable of killing her 54%* 24%
8. partner is drunk every day 42%* 12%
9. partner controls all victim’s activities 60%* 32%
10. partner beat victim while pregnant 62%* 7.7%
11. partner is violently jealous of victim (says things” if I can’t 79%* 32%
have you, no one can”)

12. victim threatened/tried to commit suicide 7% 9%

13. partner threatened/tried to commit suicide 39%* 19%
14. partner is violent toward victim’s children 7% 3%

15. partner is violent outside house 39% 37%

 

Note. * p<.05. From Campbell (2000) “Issues in risk assessment in the field of intimate
partner violence” Paper presented at the meeting of 5‘h Annual BISC-MI (Batterer
Intervention Services Coalition of Michigan) conference. C0pyright 2000 by Campbell.
Reprint by permission. Online available

0 O I O O

Regarding presence of forced sex, Campbell found that 59.5% of battered women
had been repeatedly sexually abused in the relationship and an additional 13.9% had been
raped by the batterer once. Typological studies showed that the most severe patterns
involve sexual assault and extreme sexual jealousy (Gondolf, 1988; Snyder & Fruchtman,
1981)

Substance abuse is related to criminality and recidivism in general (e. g. Harris,

Rice, & Quinsey, 1993; Monahan, 1981). Offenders with a history of family violence are

14

more likely than those with no such history to misuse substance as Dutton and Hart (1992)
and Tolman and Bennett (1990) found. However, Campbell (1986) mentioned the use of
marijuana is not considered a risk factor because of the lack of association of using this
drug with violence in the literature.

In respect to presence of threatening to kill or victim’s believing that batterer is
capable of killing her, Campbell (1986) mentioned that based on her clinical and research
experience, battered women are often the best judges of the abuser’s potential for
committing lethal violence, even though a minority of them may misjudge. According to
Campbell (2000), this item did reach significant difference as shown in Table l.

The relationship between alcohol usage and battering has been disputed in the
literature. However, Hotaling and Sugarman (1986) found that 7 of 9 (78%) studies
supported that alcohol usage is significantly associated with battering.

Regarding the presence of controlling most or all of victim’s daily activities, the
male batterers may use the various ways to control their inmate partners to ensure that
they are all being kept in line. The ways may include how much money they can spend,
when should they need to come home from working place, and how much mileage they
regular drive from working place to home etc. Campbell (1986) included the controlling

in the same category with jealousy. Both of them may be very similar in nature.

15

According to Campbell (2000), while compared attempt femicide and non-femicide
intimate assault cases, it was found that the two groups did not reach significant
difference in this item as shown in Table 2.

Concerning Battering during pregnancy, Hilberman and Munson (1977) and
Walker (1979) found that the abuse may begin during pregnancy or increase during the
prenatal period. Fagan, Steward, and Hansen (1983) reported that abuse during pregnancy
was a strong predictor of both severity of injury to the women from abuse and
extradomestic violence on the part of batterer. Furthermore, as mentioned earlier,
McFarlane, Parker, and Soeken (1995) found that women abused during pregnancy had
significantly higher scores on amount of violence in the CTS and more risk factors of
homicide in the DA, when compared with women abused prior to but not during
pregnancy. The possible reasons for including this item as a risk marker are the batterer
might be suspicious that the fetus is not his, or this assault is very dangerous to pregnant
women’s bodies by nature (Campbell, personal communication, March 24, 2000).

Regarding to presence of violent and constant jealousy, Campbell (1981) found that
male jealousy and male dominance were the reasons for homicide in 82.2% of the killing
of women by men with who they had an intimate relationship. Berk, Berk, and Loseke

(1983) also found that some evidence to support the contention that male dominance

l6

increases the severity of violence.

In respect to presence of victim’s threatening or trying to commit suicide, it was
estimated that 10% of battered women attempt to commit suicide and that approximately
26% of suicidal women seen in hospitals were also battered (Stark & F litcrafi, 1985).
Even though that correlation exists, the connection between wife battering and attempted
or actual suicide was not been specifically explored by research (Campbell, 1986).
According to Campbell (2000), while comparing attempted femicide and non-femicide
intimate assault cases, it was found that the two groups could not reach significant
difference in this item. Still, it may be worthy to trying this item in the Taiwan sample.

Regarding presence of abuser’s threatening or trying to commit suicide, Campbell
(2000) found that while comparing attempted femicide and non-femicide intimate assault
cases, it was found that the two groups did reach significant difference in this item (3 9%
versus 19%).

Regarding presence of abusing child, Monahan (1981) found that the offenders
with a history of violence are much more like to engage in future violence than those who
do not have such history. It is also found that intimate assaulters with a history of physical
or sexual assault against family members are at increased risk for violent recidivism

(Gondolf, 1988; Hotaling and Sugarman, 1986). According to Campbell (2000), while

17

 

CO

not

”gm.

S51le

comparing attempted femicide and non-femicide intimate assault cases, it was found that
the two groups could not reach significant difference in this item. Still, it may be worthy
to trying this item in the Taiwan sample.

Presence of violence outside the home did not exist in the earlier version of the DA
in 1986, but it was added to the 1995 version. Batterers with a history of violence have an
increased risk for intimate assault, even if the past violence was not toward their intimate
partner or other family members. Researchers and clinicians found that the so-called
“generalized violence people”, those who are violent both inside and outside of their
homes, engage in more frequent and more severe intimate assault than do other intimate
assaulters (Cadsky & Crawford, 1988; Fagan et a1, 1983; Gondolf, 1988; Stuart &
Campbell, 1989). According to Campbell (2000), while comparing attempt femicide and
non-femicide intimate assault cases, it was found that the two groups could not reach
significant difference in this item. Still, it may be worthy of trying this item in the Taiwan

sample.

W The Conflict Tactics Scale (CTS), developed

by Staus in 1974, is the most widely used measurement of physical violence among

cohabiting family members, including the violence between intimate partners, couples,

l8

parents-children, or siblings. The CTS is comprised of 19 items that assess tactics used in
interpersonal conflicts (see Appendix 1 for original CTS). The items are presented on a
continuum, which is from non-violence to severely violent tactics. Examples of violence
items are “threw something at partner”, “slap”, kicked, bit, or hit with a fist”, and “ used a
knife or fired a gun” etc. The CTS contains three types of conflict tactics, which are
reasoning, physical aggression, and verbal aggression. The CTS also has high internal
reliability (Straus, 1990a). Til] 1995, there were about 400 studies using the CTS. In 1996,
the CTS-2 was developed and increased to 78 items (Straus, Harnby, Boney-McCoy, and

Sugarman, 1996).

WA

The DA is one of the most commonly used risk assessment instruments for intimate
assault cases. Although the predictive validity of the DA has yet to be established, it is
one of the few instruments with any published empirical evaluation of psychometric
properties such as test-retest and internal consistency reliability. As was shown in the five
previous studies in table 3, the test-retest reliability ranged from .89 to .94. The internal
consistency was found to have alpha ranging from .60 to .86. However, as pointed out in

Campbell (1986, 1995), there is some controversy about whether internal consistency

l9

reliability is an appropriate psychometric technique to use with an instrument wherein
each item is considered to be an independent risk factor.

A study done by Campbell (2000) showed that only three items could not reach
significant difference between femicide and partner abuser controls as shown in the
following table. These three items are victim’s threatening/trying to commit suicide (Item
12), partner’s using violence toward children (Item 14), and partner’s using violence
outside house (Item 15).

Goodman, Dutton, and Bennett (2000) sampled 92 battered women from
Washington DC, who reported their cases to the court and were woulding to participate
in the study. The DA and the CTS-2 were provided to them in the initial stage, and a
three-month follow-up phone call was made to ask them whether they had been
threatened or assaulted by their partner since the arrest. Forty-nine subjects were able to
be recontacted. It was found that the DA, the CTS-2, and demographic variables did not
differ from the initial time. Mean scores on the DA and the CTS-2 were 7.24. (SD=2.56)
and 32.97 (SD=24.67). Twenty-two percent of recontacts reported their partners’
reoffense since the arrest. While checking the odds ratios obtained in the logistic
regression analysis, the authors concluded that both the DA and the CTS-2 significantly

predicted reabuse, and that the former was a much stronger predictor. When considered

20

Table 3 Reliability and Validity in Danger Assessment (DA) Research

 

 

 

 

 

 

 

 

Studies
1 2 3 4 5
N 79 abused 30 abused 52 abused 156 mixed 164
relationship
problem
Setting shelter 36% ER room Prenatal Community
64% Inpatient
OB/GYN
Ethnicity 33% minority 62% minority Black—=71 B1ack=126
White=46 Nonblack=37
Hispanic=39
Reliability Test-retest=.94 Test-retest=.89 Alpha=.86 Alpha=.66
Alpha=.60 Alpha=.67
Validity Construct(r) Construct (r) None Construct (r) Total sample(r)
CTS=5.5 Injury=.48 ISA=.75 ISA—P=.66
Injury=.50 CTS=.46 ISA-P=.44
Tactics=.43
Mean 6.3 8.7 9.2 ER room .3(non-abused) 5.5

8.3 Inpatient 3.5(abused)
OB/GYN black=2.7
white=4.4
latino=4.1
Note, From Campbell (1995). Prediction of homicide of and by battered women. In J. C.
Campbell (Ed.), Assessing dangerousness: Violence by Sexual offender, batterers, and
child abusers (pp. 96-113). Thousand Oak, CA: Sage. Copy 1995 by Campbell. Reprinted

by permission.

separately, an increment of one standard deviation on the DA was related to an
approximate four-fold (4.18) increase in likelihood of re-abuse, whereas an increment of

one standard deviation on the CTS-2 predicted only a 2.77 increase.

I 'll [1. jf']

Even though this study had defined the intimate asssaulter as shown in the previous

21

session, it is traditionally difficult to measure the battering, violence, or abuse between
intimate partners. Since this study would use the CTS as the measurement tool to reassess
the DA in Taiwan, it would be better to know how the developers of the CTS defined
violence or abuse, and what were their concerns.

WW Murry Straus, the developer of the CTS,
identified two of the most important dimensions on measuring violence—acts and
injuries. (a) Acts: The severity of the assaultive acts, by which is meant the potential for
producing an injury that requires medical treatment, and it can range from a slap to
stabbing and shooting. (b) Injury: The level of physical injury actually inflicted, which
can range from none to death (Staus, 1990c).~:

However, the conceptualization and operationalization of the abuse by the CTS is
based on the identification of certain acts as being inherently “abusive”, regardless of
whether an injury occurs. The reasons for measuring based on acts, according to Straus
(1990c), are the following six: (1) consistent with legal usage, (2) reflects humane values,
(3) injury and assault loosely linked, (4) a more realistic measure of incidence rates, (5)

psychological injury hard to measure, and (6) more useful for planning prevention

programs.

22

El'l'l' 13511 El SIS

Reliability. The internally consistent reliability of the CTS was examined using
two techniques. For Form N, the alpha coefficient of reliability was computed. These
coefficients were given in Table 4. The reliability of coefficients were high for the Verbal

Aggression and Violence scale and low for the Reasoning scale.

Table 4 Alpha Reliability Coefficients for CTS

 

 

Relationship Reasoning Verbal Physical
aggression aggression
Barling et a1 (1987) Husband-wife .50 .62 .88
Mitchell & Hodon (1983) Husband-wife -- -- .69
Straus (1987) Husband-wife .42 .77 .86
Winkler & Doherty (1983) Husband-wife .61 .81 .83

 

Note, From Straus, M. (1990a). Measuring intrafamily conflict and violence: The
Conflict Tactics (CT) Scales (p.64). In M. A. Straus & R. J. Gelles (Eds) Physical
violence in American families: Risk factors and adaptations to violence in 8, 145 families
60p.29-4 7). New Brunswick, NJ: Transaction Publishers.

Validity. Evidence of concurrent validity is reported in a study by Bulcroft and
Straus (1975), in which the CTS was completed by students in two sociology courses.
The students responded for a referent period consisting of the last year they lived at home
while in high school. They were asked to indicate how ofien during that year their father
and mother had done each of the items in the CTS. The correlations are low for the

Reasoning Scale and high for the Verbal Aggression Scale and Violence Scale (see Table

23

5). An analysis by Bulcref’t and Staus (1975) suggested that the higher correlations for the
two aggressive models of conflict are due to such acts being more dramatic and

emotionally charged and, therefore, better remembered.

Table 5 Correlation of Spousal Report CTS Scores with Student Report CTS Scores

 

Correlation (r) for:

 

 

Conflict Tactics Scale Husbands @=5 7) Wives Q1=60)
Reasoning .19 -.12
Verbal aggression .51 .43
Violence .64 .33

 

Note. From Straus, M. (1990a). Measuring intrafamily conflict and violence: The
Conflict Tactics (CT) Scales (p.41). In M. A. Straus & R. J. Gelles (Eds.) Physical
violence in American families: Risk factors and adaptations to violence in 8, 1 45 families
@p.29-4 7). New Brunswick, NJ: Transaction Publishers.

CnncismreiatedjotheCIS. There were about nine critics for the CTS as

shown in the Straus (1990b). They were (1) restricted to conflict-related violence, (2)
limited set of violent acts, (3) threats are counted as violence, (4) self-reports are
inaccurate using a one-year period, (5) equates acts that differ greatly in seriousness, (6)
conflict context is ignored, (7) ignored who initiates violence, (8) “minor” versus
“severe” classification has no empirical basis and distorts gender difference, and (9) does
not measure process and sequence.

Even though the CTS has limitations for measuring the amount of violence, it is

24

still the most widely used tool for measuring domestic violence, since it has satisfactory
reliability and validity as mentioned earlier.

Iheflimdjmmnm Tang ( 1994) explored the extent of spouse
aggression in Chinese families in Hong Kong by using the CTS. Subjects were 216
female and 136 male undergraduate students who reported on the various forms of
interparental aggression and violence. About 75% of the subjects reported interparental
verbal or symbolic aggression and 14% indicated the use of physical violence between
between parents. In general, compared to mothers, fathers engaged in more verbal
aggression against their spouses, whereas mothers were than less likely as fathers to use

actual physical force toward their spouses.

C 11] El' 1 1.1.

Even though this study did not directly focus on the causes of intimate assault, the
causes and theory of intimate assault in Taiwan are worthy of exploring.

Causesfoundjmljajwan, Some studies on the causes of intimate assault in
Taiwan were as follows.

Chen (1988) interviewed 12 abused wives and attributed the wife assault to eight

causes. First, the husbands believed in the value of patriarchy, and saw the wives as their

25

personal property, which could be treated as they wish. Second, if wife battering was one
of the traditions in the family of origin, men learned this behavior and used violence
toward their wives to deal with their anger. Third, the husbands had affairs with another
woman, and used violence toward their wives when conflict happened or for the purpose
of driving their wives to agree to divorce. Fourth, the husbands were fond of gambling or
abusing substances, and they would beat their wives when their wives could not provide
the money for them to gamble or abuse substances. Fifth, the husbands have
psychological problems, such as low self-esteem, and feel that their positions are lower
than their wives, and therefore they used violence to prove their power and authority.
Sixth, the husbands can not control their own emotions and behaviors. Seventh, the
husbands had psychiatric problems and could not control their own behavior. Last, the
husbands were military men, and they saw their wives as subordinates who can be treated
as they wish.

Tang, J. (1993) interviewed 12 battered wives and attributed the wife battering to
the following causes—~husbands’ having affairs, gambling, and alcoholism. Also the poor
relationship between wives and their mother-in-laws was also found as one of the causes.

Chou (1993) sampled 56 battered wives from battered wives supporting agencies

and asked them what they perceived as the causes of their being abused. The causes were

26

follows: They believed their husbands’ reasons were having bad moods (18.1%), drinking
(13.1%), being jealous (12.6%), job dissatisfaction (8.65%), having affairs (7.5%), losing
money in gambling (6.5%), coming home late (5.5%), and abusing drug (1.5%). They
believed the reason based on their family were wives’ having conflicts with mother-in-
laws (4.5%) and too many people in the family (2.5%). They believed the reason based
on the couple were the couple's disagreement on some issues (11.1%). They believed the _
reasons based on the wife side were they spent too much time on their own business
(2.5%) and they were neglectful toward their husbands since they were busy with chores
(1.5%).

Iheorxjnlaiwan, Straus’ (1990d) System Theory, Carlson’s (1964) Ecological
Theory, and Dobash and Dobash’s (1979) Feminist Explanatory Theory were the
examples of American theories in intimate assault, and provided good perspectives to
understanding on the intimate assault cases in the US. There was only one theory
developed in Taiwan by J o-chang Roda Chen. Chen developed a theory of spousal assault
related to cohabitation with the mother-in-law of husband, Womb Theory. Based on her
clinical experiences in treating the abused women, Chen found that the poor relationship

between wives and their mother-in-laws might trigger the occurrence of wife assault and

even cause wife abuse across the generations in these families. She theorized that the

27

reason might be that the male abusers would establish a strong but pathologically
emotional connection with their mother, while the mothers were abused by their husbands,
the abusers’ fathers. Therefore, this phenomenon was described by Chen as the mothers
using their “womb” to buildup emotional connections with their sons and prevent their
sons’ building a good relationship with their sons’ wives, because they were afraid of
losing these emotional connections and eventually losing their sons. By preventing losing
their sons, the abused mothers would tend to trigger the discord between their sons and
their daughter-in-law and eventually abuse their daughter-in-law emotionally and
psychologically, or trigger their sons to physically abuse their wives. This phenomenon
would pass on across the generations in these families, since the sons’ wives would
eventually replicate these patterns in how to treat their daughter-in-laws (Chen, 1997).
This theory would provide a good understanding for the wife assault specifically in
Chinese culture, since the Chinese culture traditionally encourages the sons having the
obligation to live with their old parents. However, less and less daughter-in-laws in
Taiwan would like to live with the husbands’ old parents. Therefore, the Womb Theory
might decrease in its ability to explain the discord among couples who did not live with

the husbands’ old parents.

28

11' El'S '1’

Taiwan has a population of 230 million, most of which lives in the Taiwan island
with an area of 36,000 square kilometers (see Figure 1). Taiwan’s population can be
usually divided into four ethnic groups, Ming-nan, Hakka, mainlander, and aboriginal
(Chen, 2000).

There are two groups of early Chinese immigrants. The first of them are the
Fuchienese (also known as “Ming-nan”—-means south of Fuchien Province in China),
came from southern China’s Fuchien Province, which is directly across the Taiwan Strait
400 years ago. They comprise 70 percent of the Taiwan population.

The second group of Chinese immigrants were the the Hakka. The Hakka came
from south China shortly afier the Ming-nan immigrated to Taiwan. The population of
Hakka in Taiwan is about 3 million, which is 15% of Taiwan population. Today there are
still Hakka-dominated areas in Taiwan, though this group of Chinese can also be found
throughout Taiwan. They speak the Hakka dialect in addition to regular Taiwanese (a
derivative of the Fuchien dialect) and Mandarin. Hakkas are in a number of ways

culturally distinct from other Chinese people (Chu, 1998).

29

Map of Taiwan

Figure 1

«15¢...» a

 

 

 

1n

a

IS

i

C

rung

ainland China (2) Kaohs

southwestern Taiwan. Kaohsiunghsien, which is around Kaohsiung City, means

Kaohsiung County.

theM

18

Note. (1) The land. at the lefi edge

30

The third group is comprised of Chinese from various parts of China who came to
Taiwan after World War II, mostly in 1949, the time after the retreat of Kuomingtan
(KMT, the ruling party of that time ruled by Chiang, Kai—shek). They comprise just under
15 percent of Taiwan population. They are referred to as “mainlanders”.

The aboriginals or earliest inhabitants consist of 9 mountain aboriginal tribes and 9
plain aboriginal tribes. Most of the latter tribes have been civilized by the mainstream
Chinese culture, and are hard to identify as their own cultures. The following would
mainly focus on the mountain aboriginals. The mountain aboriginals are considered to be
of Malay or Polynesian origin based on their languages and cultures. They comprise less
than 2 percent of the Taiwan population (about 450 thousand people). There are 9
different mountain aboriginal tribes. Although the aboriginals speak languages that
resemble Malay, rather than Chinese, they do not have a common language among them
other than Japanese or Chinese. Many still live by hunting and fishing, but aboriginals are
now working in other occupations especially those associated with tourism. They are

generally poorer than the rest of population and their birth rate is lower.

[IV C 1 El 1 . . I 'w
Whit; As mentioned earlier, about 98% of Taiwanese population

31

are from Mainland China in a different time, so Taiwanese peOple have been keeping the

traditional Chinese Culture, especially Confucianism, which was created by Confucius at

about 2500 years ago and has been most influencing Chinese culture in the past 2000

years. Confucianism emphasizes the following values related to the family:

(1) F ilial piety: There is a Chinese saying, “F ilial piety is the first of hundreds of
benevolence”. It is one of the traditional Chinese cultural beliefs, which the Taiwan
people still keep. The son as well as his wife needs to take good care of the parents
while the parents are elderly, not to mentioned live together with the parents.
Moreover, filial piety is the most basic requirement for a person to earn dignity from
his community and government.

(2) Familial collectivism: The basic unit of ethnicity in traditional Chinese culture is seen
as family (or even clan), instead of individual (Lee, 1978). Therefore, an individual
needs to take good care of the family’s reputation. Accordingly, the clan can be also
viewed as one of the social control components, in addition to parents.

(3) Shame culture: Bendict (1946, cited from Chin, 1990, p322), an American
anthropologist, identified that Chinese culture is a shame culture, whereas Western
Culture is a guilty (sin) culture. Shame culture encourages accepted behaviors by

external sanction, which is more of an informal social control and then intemalizes it

32

to personal beliefs. Moreover, Chin (1990) proposed that shame comes from the
concept of “1E3. LI” (courtesy), which means routinized courtesy in family, groups,
school, and government in Chinese culture. Everyone needs to follow “L1”. If
someone did not follow it, that would hurt the reputation not only of the individuals
but also of the family or clan.

Gallin (1999) did a sociological review of 25 married women at a southern Taiwan
village in the early 805 and found that only one reported to be treated well by her mother-
in-law, whereas 12 of them reported to be bruised by the mother-in-law. She described the
following, which are very close to the real situations in Taiwan.

0 China’s partrilineal kinship structure recognizes only male children as descent group
members with rights to the family’s property. In the past, and to a large extent today,
residence was patrilocal.

0 Females were considered as a liability—a household member who drained family
resources when in childhood, whereas she needs to leave her natal home to live as a
member of her husband’s family, cutting her formal ties with her family of origin.

0 In Chinese tradition, filial piety (Shiao) obligates the offspring to repay parents for
nurturing them. Therefore, the wives are required to be subordinate to the parents-in-

law.

33

o A young wife had to be socialized and integrated into the household, and in the
division of labor this task fell to her mother-in-law. To enforce obedience, and
perhaps to retaliate for her own lifelong subjugation, an older woman might well
resort to physical abuse of a daughter-in—law with whom she was dissatisfied.

0 Ten of the 25 married women reported that men also had beaten their wives to keep
them in line. These men had acted as agents of their mothers, who had demanded

they bend their wives to the older women’s willingness.

It revealed that the relationship between wife and mother-in-law in Taiwan may be
generally poorer that in the US. This situation can be one of the risk markers in Taiwan
and deserves to be addressed in this study. However, as Campbell (1999) pointed out,
wife battering may be, if anything, decreasing as extending family living arrangement
become less common. As the researcher know, this has resulted from younger couples
tending to live in the cities for work whereas the parents-in-laws like to live in rural areas,
and also, young wives generally dislike living with the parents-in~laws.

Hakkaflilture. The Hakka keep traditional Chinese Culture, but it is generally
recognized that there are differences between Hakka and other Chinese such as “A close-

knit people with their own distinctive customs and dialect, the Hakka have never been

34

completely assimilated into Chinese society—they maintained their cultural solidarity,
intermarrying rarely with other Chinese...” (American Encyclopedia, 1991). Moreover,
the Hakka are an extremely industrious, shrewd peOple, and tend to be very clannish as
well. However, even though the above situations may change more or less in the modern
era, the Hakka may still keep most of the above cultural beliefs to some extent.
Aboriginalflulmre. There are 9 aboriginal tribes in Taiwan, and each of them
have the different cultures and distinct languages. The 9 tribes are Ami, Atayal, Paiwan,
Bunun, Rukai, Puyurna, Saisiyat, Tsao, and Yami. According to Hung (1993), the
populations of each are 123000, 78000, 59700, 38000, 8000, 8000, 43000, and 4200,
respectively. All of the aboriginal tribes still keep the hierarchy of their social class. Ami,
Saisiyat, and Yami are maternal societies, which means the mother has the authority over
the families and their property. In the past five years, they have been suffering from a
high unemployment rate (about 15% or more males), since the government induces the

cheaper immigrant laborers for the industry, instead of employing them.

11' El" 'I'w

As the synthetical study of Hotaling and Sugarman (1986) showed earlier, religious

incompatibility was found in 100% of previous studies, including religious

35

incompatibility supported as being a risk marker of wife assault. Therefore, it would be
worthy of exploring the diverse religions in Taiwan.

In terms of religions or religious beliefs, the Taiwanese people are quite eclectic
and most people report adhering to more than one religion, except Christianity and
Catholicism. The Majority of Taiwanese people, 65 percent according to a recent poll,
believe in Taiwan folk religion, which can be seen in the southern division of Taoism in
China. But, since Chinese immigrants also brought Confucianism, Taoism, and Buddhism
to Taiwan, the majority of Taiwan peOple report being followers of one or all of these
religions. The aboriginals do natural worship and various sacrifices. About 7% of

population believe in Christianity or Catholicism (Chen, 2000).

Some differences on the dynamics of intimate violence between Taiwan and the
US can be organized as follows.

Taiwan intimate violence cases might not have the same “honeymoon or loving
respite phase” as the US. has. Walker (1979) studied battered women and found that
there is a “cycle of violence” in intimate violence. There are three phases in this cycle of

violence. First, there is tension building. In this phase, minor incidents of violence may

36

occur along with a buildup of anger. This phase may include verbal put-downs, jealousy,
threats, and breaking things and can eventually escalate to the second phase. Second, in
the acute or battering phase, major violent outbursts occur. This violence can be seen as
the major earthquake. Following the second phase, the couple sometimes enters phase 3.
Third, in the honeymoon or loving respite phase, the batterer is remorseful and afraid of
losing his partner. He may promise anything, beg forgiveness, buy gifts, and basically be
“the man she fell in love with.” However, according to Tang and Tsai (1997), the majority
of Taiwan abused women would not go to the shelters, and, instead, they would go to the
family of origin or house of their own siblings. Then, the typical abusive men would
intimidate the wives’ relatives to let go or even plead with their wives to come back.
After their wives come back with them, the abuse continues. Therefore, Tang and Tsai
(1997) concluded that it may not necessarily have a “honeymoon and respite phase” in
Taiwan cases, and instead, the abuse happens as a cycle of escape-and-violence inside or
outside the house.

Furthermore, Tang and Tsai (1997) pointed out that some Taiwan intimate abusers
may also aim their violence toward elderly persons, such as their own parents or parents-
in-law, who live in the same house. However, this might not be the case for the US.

families, since most of American families do not live with their elderly parents.

37

Based on their cases, Tang and Tsai (1997) pointed out that physical or
psychological victimization in childhood was not always found in the Taiwan intimate
abusers’ history, and instead, they found some of them were “spoiled” by their parents.
Hotaling and Sugarman (1986) synthesized 52 previous studies and stated that while not a
consistent risk marker, experiencing violence fi'om parents/caregivers has been found to
discriminate batterers and nonbatterers in the majority of studies.

Regarding firearm to be used in the intimate assault, Hart (1988) and Straus (1991)
indicated that the presence of weapons is one of the life-threatening risk factors among
intimate abusers. Moreover, the Danger Assessment scale (Campbell, 1995) also included
this factor in assessing the lethality among intimate assaults. However, since Taiwan is a
firearrn—control country, the people are not permitted to own guns, except for special
permission to own a rifle for the purpose of hunting or shooting practice for competitions.
Therefore, this risk factor for lethality may not fit for Taiwan intimate assault cases.

Poor relationship between wife and mother-in-law could encourage wife assault in
Taiwan. Since wives in Taiwan must be obedient to their spouse and parents-in-law,
Taiwanese women may be abused to be kept in line.

Feminist theory may not totally fit in Taiwanese wife assault cases. According to

Campbell (1999, p267) both cross-cultural and individual studies make it clear that there

38

is not a simple linear correlation between female status in a society or in a couple and
rates of wife assault (Campbell, 1985). In fact, there may be a curvilinear relationship,
with wife battering being relatively low in societies where women’s status is the lowest,
since there are other norms that effectively control women (Campbell, 1985). Women’s
status and power are complex variables with many different manifestations. On the other
hand, theoretically when women have power in the varied cultures outside of the home,
either economic (Wape, Mayotte etc. the culture names) or magical (Kalai and Ujelang
etc), it does seem to provide some protection against wife beating that supports of the
feminist case. It can be true in some cultures, but seems not to be in Taiwan. There is
some negative evidence in Taiwan where societal development has afforded economic
possibilities for wives but battering continues (Gallin, 1999). Campbell (1999) postulated
that the strong continuing ethic of females as property and the societal norms supporting
the beating of wives in those cultures is so far overcoming any development of autonomy

in women of Taiwan.

DV] [B'll [If] le
Risk assessment has been done for over centuries. However, based on Bonta (1996),

it was not until 1928 that scientists started to develop risk assessment tools based on

39

scientific methods. Bonta (1996) called the risk assessment done by 1928 first-generation
risk assessment, which was done using the subjectivity of professionals. The most serious
weakness of this approach is that the rules of collecting information and formulating
interpretations of the data are subject to considerable personal discretion, and make
accountability and fairness difficult. The second-generation risk assessments are the
objective and empirical risk assessment. This kind of risk assessment can be tracked back
to Burgess’s (1928) study of over 3,000 parolees. Burgess identified 21 factors that
differentiated parole successes from parole failures and he used these factors to construct
a risk scale. Then Glueck and Glueck (1950) also developed a risk assessment scale for
delinquent behavior based on the variables they found that could differentiate the
delinquents from the nondelinquents. However, their major weakness, based on Bonta
(1996), was that the instruments provided little direction for treatment and most of them
only contained the historical/static factors such as criminal history. Finally, Bonta ( 1996)
contended that the third-generation risk assessment has to be developed for the purpose of
connecting the treatment implication based on the risk assessment. He developed the
concept of criminogenic need, which is the need that cause criminals to commit crimes
and generalized the criminal behavior. Criminogenic need is a dynamic factor, and the

likelihood of committing crimes could change if the criminogenic need is altered.

40

Moreover, in 1994, Monahan and Steadman (1994) called for a new generation of
risk assessment, which should be based on the actuarial method. Monahan et a1 (1994)
noted that actuarial assessment observes the relationship between specific cues or risk
factors and the occurrence of violent behavior.

Generally predictors of convenience are rarely those that would be chosen on
theoretical grounds. However, Quinsey, Harris, Rice, and Corrnier (1998) submitted an
example that if the prediction based on a theory of horse race performance related the
characteristics of the horses (such as lung capacity or training method) while predicting
the winner in a horse race, it would improve understanding more than an atheoretical
actuarial approach. Therefore, Quinsey et a1 (1998) concluded that there is no reason why
theoretically relevant predictors could not be used in an actuarial model. Therefore, there

is no concrete theoretical approach for identifying risk marker.

I l . [I l l l . : 1
After reviewing the previous typologies of male batterers by means of either

rational-deductive (9 studies) or empirical-inductive (6 studies) methodology,

Holtzworth-Munroe and Stuart (1994) identified three dimensions to distinguish the

different types of male batterers. The three dimensions were severity of marital violence,

41

generality of the violence (toward the wife only or toward the others also), and
psychopathology/personality disorders. Afier this identification, they proposed a typology
consisting of 3 subtypes of male batterers, which were family only batterers,
dysphoric/borderline batterers, and generally violent/antisocial batterers. Accordingly,
they presented a development model of marital violence to theoretically explain how the
three subtypes of male batterers differed in this model. First, family only batterers
engaged in the least severe marital violence and are the least likely to engage in
psychological and sexual abuse. The violence of this group is generally restricted to
family members; these men are the least likely to engage in violence outside the home or
to have related legal problems. They show little psychopathology and either no
personality disorder or a passive-dependent personality disorder. Second,
dysphoric/borderline batterers are found to engage in moderate to severe wife abuse,
including psychological and sexual abuse. This group’s violence is primarily confined to
family, although some extrafamilial violence and criminal behavior may be presented.
These men are the most dysphoric, psychological distressed, and emotionally volatile.
They may evidence borderline and schizoidal personality disorder characteristics and
may have problems with alcohol and drug abuse. Third, generally violent/antisocial

batterers engage in moderate to severe marital violence, including psychological and

42

sexual abuse. These men should engage in the most extrafamilial aggression and have the
most extensive history of related criminal behavior and legal involvement. They are
likely to have problems with alcohol and drug abuse, and they are the most likely to have
an antisocial disorder or psychopath.

In 2000, Holtzworth-Munroe and her colleagues had tested this proposed typology
on a sample of 102 male batterers by cluster analysis, and found that in addition to the
three subtypes, a fourth subtype was identified and named as “low-level antisocial
batterers”, which were between family only batterers and generally violent batterers

(Holtzworth-Munroe, Meehan, Herron, Rehman, & Stuart, in press).

43

CHAPTER 3 METHOD

Hypotheses

Because this study introduced the Danger Assessment scale into Taiwan and
because the author found some differences in the literature review, it might be necessary
to consider the cultural difference between Taiwan and America. The hypotheses of this
study were as follows.

1. It was hypothesized that the Danger Assessment scale could significantly predict the
amount of violence measured by the Brief Conflict Tactics Scale among the intimate
violence cases in Taiwan. The Danger Assessment scale included increased fi'equency
of physical violence, increased severity of physical violence, occurrence of choking,
presence of firearms, occurrence of forced intercourse, drugs usage, threatening to kill
victim or victim’s perception of capability to kill, drinking, controlling daily behavior
of victims, physical assault while pregnant, violent jealousy, suicidal behavior or
attempt of both parties, violence toward children, and violence outside the house. Item
4 (owning a gun) would be excluded, since it asks whether people have a gun at home

or not, and Taiwan is a firearm-control country and accessing to a gun is relatively

44

difficult in Taiwan. The Brief Conflict Tactics Scale used to measure the amount of
violence, included “throwing object”, “pushing, grabbing, or shoving”, “slapping”,
“kicking, biting, or hitting with a fist”, “hitting or trying to hit with something”,
“beating up”, “choking”, “threatening with a knife”, and “using a knife and any other
killing behaviors except choking”, in all 9 previously mentioned physically violent
behaviors.

2. It was hypothesized that the Danger Assessment scale could significantly predict the
lethal violence among the intimate violence cases in Taiwan.

3. It was hypothesized that other variables, such as victims’ perception of future abuse,
abusers’ stalking behavior, abusers’ threatening to hurt the victims’ family of origin,
and abusers’ threatening to hurt the children, which are not included in the Danger
Assessment scale could also significantly predict the amount of intimate violence as
well as lethal violence of the abusers.

4. Finally it was hypothesized that DA could be used to identify traits that distinguish

different types of male intimate abusers.

This study administered the Brief CTS to the female victims to measure the amount

45

of violence their intimate partners inflicted on them. Originally, there are 19 items in the
CTS-Form R (see Appendix 1), and all of them are in the three categories. Item 1 through
Item 3 (“discussed an issue calmly”, “got information to back up his side of things”, and
“brought in, or tried to bring in, someone to help settle things”) belong to the “reasoning”
category, Item 4 through Item 10 (“insulted or swore at you”, “sulked or refused to talk
about an issue”, “stomped out of the room or house or yard”, “cried”, “did or said
something to spite you”, “threatened to hit or throw something at you”, and “threw or
smashed or hit or kicked something”) belong to the “verbal aggression” category, and
Item 11 through Item 19 (“threw something at you”, “pushed, grabbed, or shoved you”,
“slapped you”, “kicked, bit, or hit you with a fist”, “hit or tried to hit you with
something”, “beat you up”, “choked you”, and “threatened you with a knife”) belong to
the “physical aggression” category. Since the nature of the DA scale is to focus on
physical aggression, this study would only focus on physical aggression. Therefore, the
researcher only included Item 11 through Item 19 (“threw something at you”, “pushed,
grabbed, or shoved you”, “slapped you”, “kicked, bit, or hit you with a fist”, “hit or tried
to hit you with something”, “beat you up”, “choked you”, and “threatened you with a
knife”) of the original CTS for a total of 9 items on the Brief CTS. Then this study would

follow the following rationales to score the BriefCTS:

46

1. Item 1 through Item 3 (“threw something at you”, “pushed, grabbed, or shoved you”,
and “slapped you”) can be viewed as “minor violence”, whereas Item 4 through Item 9
(“kicked, bit, or hit you with a fist”, “hit or tried to hit you with something”, “beat you
up”, “choked you”, and “threatened you with a knife”) can be viewed as “severe
violence” (Straus 1990a).

2. Scoring in accordance with the frequencies: In the Brief CTS, the answer of 0
represented never in the past year, 1 represented 1 time in the past year, 2 represented 2
times in the past year, 3 represented 3 to 5 times in the past years, 4 represented 6 to 10
times in the past year, 5 represented 11 to 20 times in the past year, and 6 represented
over 20 times in the past year of a particular violent behavior (Straus, 1990a).

3. According to Straus (1990e), the weighting methods for each violent behavior in the
Brief CTS are to multiply the frequency of each violent behavior by the following
weights. The weights for each item are “throwing something”=l , “pushing, grabbing,
or shoving”=l , “slapping”=1, “kicking, biting, or hitting a fist”=2, “hitting or trying to
hit with something”=3, “beating up”=5, “choking”=5, “threatening with a knife”=6,
and “using a knife”=8.

The Danger Assessment (DA) scale was translated to a Chinese version and

provided to the victims in this study. However, Part A of the DA scale, a part requiring

47

the victims respond on the times of being abused in each month of the previous year,
would be excluded, because this study did not used chronicled data. Moreover, Item 4
(owning a gun), would be excluded, since Taiwan is a firearm-control country and
civilians have very limited chances to access firearms.

In addition, the researcher created a Victim Questionnaire for female victims for
the purpose of collecting demographic information and testing some other risk markers
supported by previous researches, such as cultural and religious difierences between
intimate partners, employment status, income level, educational level, victims’ self—

assessing on danger, and being stalked etc.

WWW

Samplingpmsess. Kaohsiung City and Kaohsiung County are adjacent each
other in southern Taiwan. Both areas were reported as the two highest areas of intimate
assault in Taiwan based on the reported case amount of local Domestic Violence
Prevention Centers, even though both areas are not the city or county with the highest
population in Taiwan. Kaohsiung City, with a population of 1.4 million, is the second
biggest city in Taiwan, whereas Taipei City is the biggest city in Taiwan with a

population of 2.6 million. Kaohsiung County, with a population of 1.2 million, is the 7th

48

biggest local area (including cities and counties), whereas Taipei County, with a
population of 3.5 million, is the biggest local area in Taiwan (Taiwan Ministry of Interior,
2000). According to telephone interviews with the directors of the Domestic Violence
Centers in Kaohsiung City and Kaohsiung County, the monthly case amount of intimate
violence is 250 and 230, respectively, in the 24-hour hotline, and 95 and 115,respectively,
would be new-open cases in both agencies, based on the victims’ willingness for further
help and the seriousness of the assault (personal communication, May 22, 2000). The
sampling process of this study would try to collect the intimate abuse cases from the two
government agencies, because they can continuously meet with the victim and learn more
information about their families and their partners. However, it needs to be noted that this
resource could be biased, since it may automatically be screened by the severity and the
nature of the intimate abuse (such as the parents-in-law may not want victim to report the
violence or they may want the victim to drop the case). The time period for data
collecting would be from December 17, 2000 to January 12, 2001. In total, there would

be four weeks for the data collection in this study.

W The DA, the Brief CTS, and the Victim Questionnaire

would be provided to victim when the victim sought further service and the social

49

workers in Domestic Violence Centers started the intake process. Social workers were
trained how to administrate the instruments skillfully and ethnically. The participants of
this study were treated based on ethical guideline of American Psychological Association
(APA). (see Consent Form for respondents and Ethical Response Sheet for Social

Workers in Appendix 5 and Appendix 6, respectively).

1.! .-. u ' or ‘ 0 ‘ an. {11.1. ' L at not 1, 3, Ln ‘ r
In this study, two research instruments were used. Conflict Tactics Scale (CTS) was
simplified to Brief CTS as of only measuring the severe violent behaviors. Danger
Assessment (DA) remained almost the same, except for deleting whether there is a

firearm in the house (Item 4).

Imnslafionflmceduresrrffiesearchlnstmments. The author was granted the

permission of translating and using the Brief Conflict Tactics Scale and the Danger
Assessment in this study by their original developers, Murry Straus and Jacquelyn
Campbell, respectively. After the author’s translation, four Taiwan Ph. D. Students at
Michigan State University (majors in Educational Measurement, English, Counseling
Psychology, and Sociology) were invited to check the content validity and face validity of

the translated instruments. They graded each item of the two translated instruments on a

50

four-level scale based on how appropriate and how fluent each item was translated. In the
grading, four meant very appropriate or fluent, three meant appropriate or fluent, two
meant inappropriate or not fluent and change is necessary, and one meant very
inappropriate or not fluent at all and deletion is necessary. The author changed some
items of the two translated instruments based on their grading and suggestions on how to
make appropriate translations.

Refiabflitylesmffimslatedfiesearchjnsmm After checking content
validity and face validity, the author requested the social workers in Kaohsiung City
Domestic Violence Prevention Center and Kaohsiung County Domestic Violence
Prevention Center administer the instruments to a total of 30 cases from both Center to
test the reliability. Eleven cases were collected in Kaohsiung County and 19 cases were
collected in Kaohsiung City. The Cronbach’s alpha was used for testing the reliability
based on the total of 30 cases. It was noted that Cronbach’s alpha is a model of internal
consistency, based on the average inter-item correlation (SPSS, 1999a).

(1) Danger Assessment scale: The reliability test of translated Danger Assessment was
shown in Table 6. It was found that Cronbach’s alpha is .73, which is in the
satisfactory range of such measurement. Usually the satisfactory minimum in

corrected item-total correlation is .20 and it was found that Item 6 (.14), Item 1 (.19)

51

and Item 12 (.19) are slightly under the linrit, but the author still decided to keep these

three items, since they are not significantly below.

Table 6 Reliability Test of Chinese Version Danger Assessment

 

 

Scale mean Scale Corrected Squared Alpha if
if item variance if item-total multiple item
deleted item deleted correlation correlation deleted

Item

da_l increase frequency 5.80 9.13 [E] .46 .73
da_2 increase severity 6.10 8.09 .52 .43 .69
da_3 choking 6.07 8.69 .30 .54 .72
da_5 forced sex 5.83 8.42 .45 .60 .70
da_6 drug abuse 6.40 9.42 [321] .40 .73
da.7 threat to kill or

believe to be killed 5.87 8.26 .49 .66 .70

da_8 alcohol abuse 6.00 7.79 .64 .72 .68
da_9 control daily life 5.93 8.90 .24 .40 .73
da_lO assault when pregnant 6.17 8.98 .22 .61 .73
da_ll violent jealousy 6.03 8.17 .49 .50 .70
da‘12 she want suicide 6.20 9.06 [E .54 .73
da_l3 he wanted suicide 6.33 9.06 .25 .45 .72
da_14 violence to child 6.03 8.52 .36 .55 .71
da_15 violent outside 6.17 8.83 .26 .29 .72

 

Note. (1) Cronbach’s alpha = .73; (2) Item 4 (owning the gun) was excluded, because

Taiwan is a firearm control country and people have very limited access to gun.

(2) Brief Conflict Tactics Scale: The reliability test of the translated Brief Conflict Tactics

Scale was shown in Table 7. It was found that Cronbach’s alpha was .90, which was

in the satisfactory range of such measurement. Moreover, not any one item-total

correlation in the Brief Conflict Tactics Scale was below .20, which meant all items

were in the acceptable range under this reliability test. Therefore, all items of Brief

Conflict Tactics Scale were kept in the study.

52

Table 7 Reliability Test of Chinese Version Brief Conflict Tactics Scale

 

 

Scale mean if Scale Corrected Alpha if
item deleted variance if item—total item
item deleted correlation deleted

Item
CFS_1 throw object 15.57 98.33 .59 .89
C1‘S_2 push, grab, or shove 15.36 103.28 .75 .88
crs_3 slapping 16. 61 98. 32 .66 .88
CIS__4 kick, bit, or hit with a fist 15.54 95.96 .79 .87
Cl‘S_5 hit or try to hit with something 15 .90 91.95 .85 .87
CIS_6 beat up 16.18 93.93 .83 .87
CI‘S_7 choke 17.39 100.32 .65 .88
CI‘S_8 threat withaknife 18.11 111.65 .56 .89
CTS_9 use a knife 18. 50 122.63 .24 .91

 

Note, (1) Cronbach’s alpha = .90; (2) Number of cases was 28, since two cases missed.

Measurement

1. Multiple Regression

(1) Dependent variable: The dependent variable would be from the Brief CTS outcome.

The researcher would sum up all of the 9 items after weighting each item, and use

this total as the dependent variable for multiple regression. However, since it was

found that the results from the scale was extremely skewed, which was the poisson

distribution, so it would be better to do the poisson regression for dealing with the

extreme outliers. The author would use STATA 6.0 to accomplish this task. Other

than poisson regression, the author would use SPSS 10.0 to accomplish most of the

statistical works in this study.

(2) Independent variables: The independent variables would be the risk markers in the

53

DA and the risk markers in the Victim Questionnaire supported by previous studies,
such as victims’ self assessing danger, male partners’ stalking behavior, etc.

2. Receiver Operative Characteristics (ROC) curve: The author would use the ROC curve
to estimate the predictive accuracy on the DA and the Brief CTS’s prediction toward
the appearance of lethal violence.

3. Principle Component Analysis and Cluster Analysis: The author would use the
principle component analysis to identify the key variable(s) in the DA, and try to use
it/them to distinguish the subtype of male intimate abusers. Moreover, the author
would use the cluster analysis to test whether the key variable(s) could be the criterion

(criteria) to distinguish the subtypes of male intimate abusers.

54

 

CHAPTER 4 RESULTS

Demographic Analysis

Samplesiruhemdareas, The author set a time period for social workers in the
two areas to collect data from the intimate violence victims. The time period was set as
from December 17, 2000 through January 12, 2001, totally 27 days. During this period,
122 interviews were conducted. However, one was excluded because of not enough
response in the questionnaire and scales. Therefore, 121 respondents would be counted in

this study (Table 8).

Table 8 Populations, Numbers of Monthly/Yearly Cases, and Numbers of Samples in
Kaohsiung City and Kaohsiung County

 

 

 

 

 

 

\Areas Kaohsiung City Kaohsiung County
Population (December 2000) 1,490,560 1,234,707
Number of monthly/yearly 250/3000 173/2081
cases(Year 2000)

Number of samples in this 82/135 (60.7%) 39/210 (18.6%)
study/number of case (12/17/2000~1/12/2001) (12/17/2000~1/19/2001)
happened 121

 

 

 

Note, (1) The data collecting period in Kaohsiung County was extended 7 more days,
since the local social workers had tight schedules with other social welfare services
during the time. (2) Populations in both areas were from Taiwan Ministry of Interior
(2001).

e v' ' l . In the female victims, it was

55

found that the average age of victims in this study was 35.44. Regarding ethnicity, 77.5%
were Min-nan, 8.3% were Hahkka, 5% were 2Ml generation mainlander, and 4.2% were
aboriginal. It was noted that 2.5% were from China, and 2.5% were foreigners (two from
Indonesia and one from America). Regarding educational level of total victims, it was
found 1.7% were illiterate, 20.8% were in elementary school level, 20.6% were in junior
high school level, 47.5% were in senior high school level, 3.3% were in junior college
level, 5.8% were in 4-year college level, 0.8% were in graduate school level. Regarding
religion of total victims, it was found that 29.8% had no religion, 40.5% believed in
Buddhism, 21.5% believed in Taoism, 2.5% believed in Onelawism, 4.1% believed in
Christianity, none believed in Catholicism, 1.7% believed in other religions. Regarding
employment status of total victim, it was found that 38.7% were in housewife or
unemployment status, 52.1% were in full-time employment status, and 9.2% were in part-
time employment status. Regarding the marital status of all subjects, 88.3% were married,
6.7% were divorced but cohabited, and 5.0% were cohabited without marriage (Table 9).
W In the male abusers, it was found
that the average age of male abusers in this study was 39.86. Regarding ethnicity, 70.1%
were Min-nan, 10.3% were Hahkka, 0.9% were 1“l generation mainlander, 12.8% were 2nd

generation mainlander, and 3.4% were aboriginal. It was noted that 2.5% were from

56

China, and 2.5% were foreigners (two from Indonesia and one from America). Regarding
educational level of total abusers, it was found 0.8% were illiterate, 13.6% were in
elementary school level, 26.3% were in junior high school level, 45.8% were in senior
high school level, 8.5% were in junior college level, 4.2% were in 4-year college level,
0.8% were in graduate school level. Regarding religion of total abusers, it was found that
41.0% had no religion, 23.1% believed in Buddhism, 26.5% believed in Taoism, 1.7%
believed in Onelawism, 6.0% believed in Christianity, and 1.7% believed in Catholicism.
Regarding employment status of total abusers, it was found that 20.2% were in
unemployment status, 67.5% were in full-time employment status, and 12.3% were in

part-time employment status (Table 10).

57

Table 9 Demographic Descriptions of Female Victim Respondents in This Study

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Kaohsiung Kaohsiung Total
City CountL
Number of sargrle 81 39 121
Age 36.93 36.71 35.44
Ethnicity
Min-nan 62(76.5%) 3 1 (79.5%) 93(77.5%)
Hahkka 7(8.6%) 3(7.7%) 10(8.3%L
1st gen mainlander 0 0 0
2':an mainlander 3(3.7%) 3(7.7%) 6(5.0%)
3"l gen mainlander 0 0 0
Aboriginal 34(4.9%) 1(2.6%) 5(4.2%)
From China 3(3.7%) 0 3(2.5%)
Foreigner 2(2.5%) 1(2.6%) 3(2.5%)
Educational level
Illiterate 0 2(5. 1%) 2(1 .7%)
Elementary school 17(21.0%) 7(17.9‘VQ 24(20.8%)
Junior high school 17(21.0%) 8(20.5%) 25(47.5%)
Senior high school 35(43.2%) 22(56.4%) 57(47.5%)
Junior college 4(4.9%) 0 4(3.3%)
4-year college 7(8.6%) 0 7(5.8%)
Graduate school 1(1.2%) 0 1(0.8%)
Religion
None 23(28.0%) l3(33.3%) 36(29.8%)
Buddhism 3 5(42.7%) 14(35.9%) 49(40.5%)
Taoism 18(22.0%) 8(20.5%) 26(21 .5%)
Onelawism 1(1.2%) 22(56.4%) 3(2.5%)
Christianity 5(6.1%) 0 5(4.1%)
Catholic 0 0 0
Other 0 2(5.]%) 2(1.7%)
Employment status
Housewife/unemployed 3 1 (38.3%) 1 5( 39.5%) 46(3 8.7%)
F ull-time job 43(53.0%) 19(50.0%) 62(52.1%) 73(61.3%)
Part-time job 7(18.6%) 4(10.5%) 11(9.2%)
Marital status
Married 70(85.4%) 36(94.7%) 106(88.3%)
Divorced but cohabited 7(8.5%) 1(2.6%) 8(6.7%)
Cohabited w/o married 5(6.1%) 1(2.6%) 6(5.0%)

 

 

58

 

Table 10 Demographic Descriptions of Male Abuser Respondents in This Study

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Kaohsiung City Kaohsiung Total
County
Number of sample 81 39 121
Age 40.56 38.37 39.86
Ethnicity
Min-nan 63@.8%) 19(51.4%) 82(70.1%)
Hahkka 3(3.8%) 9(24.3%) 12(10.3%)
lSt gen mainlander 1(1.3%) 0 1(.9%)
2"‘1 gen mainlander 8(10.0%) 7(18.9%) 15(12.8%)
3rd gen mainlander 3(3.8%) 1Q.7%) 0
Aboriginal 2(2.5‘V9 1(2.7%) 4(3 .4%)
From China 0 0 0
Foreigner 0 0 0
Educational level
Illiterate 0 1 (2.6%) 1 (8%)
Elementary school 12(15.0%) 4Q0.5%) 16(13.6%)
Junior high school 23(2_8.8%) 8(21.1%) 31(26.3%)
Senior high school 33(41.3%) 21(55.3%) 54(45.8%)
Junior college 7(8.8%) 3Q.9%) 10(8.5%)
4-year college 4(5.0%) 1Q.6%) 5(4.2%)
Graduate school 1(1.3%) 0 1(.8%)
Religion
None 33(40.7%) 15(41.7%) 48(41.0%)
Buddhism 21(25.9%L 6(1 6.7%) 27(23.1%)
Taoism 16(19.8%) 15fi1.7%) 31(26.5%)
Onelawism 2(2.5%) 0 2(1.7%)
Christianity 7(8.6%) 0 7(6.0%)
Catholic 2(2.5%) 0 2(1.7%)
Other 0 0 0
Occupation
F ull-time job 49(64.5%) 28:7 3.7%) 77(67.5%) 91 (79.8%)
Part-time job 8@.5%) 6Q5.8%) 14(12.3%)
unemployed 19(25 .5%) 4(10.5%) 23(20.2%)

 

59

 

5 IE. 1. 'E? lE'ECTSSr
QeneraLfindjngjnDAscores. There were two changes in the translated DA

version. The original Item 6, “partner threatened to kill victim or victim believes partner
is capable of killing her” was divided into two items. These were “partner threatened to
kill victim”(ltem 6(1)) and “victim believes partner is capable of killing her”atem 6(2)),
respectively, since the former was objective truth and the latter was personal subjective
belief suggested by local social workers after doing the pretest. Another change was the
exclusion of the original Item 4 “a gun is present in the house”. The reason for this
exclusion was that Taiwan is a firearm-control country and people have very limited
access to own firearms. Because of these two changes, there might be problems to
compare the DA in this study to the American studies.

It was found that the mean in DA was 6.04 (with standard deviation of 3.01 ), which
was close to the mean of 6.3 in Study 1 of Table 3 (a study sampling from general abuse
cases, instead of from shelter and emergency room or impatient settings). However, it had
to be cautious in interpreting any meaning because the DA versions used in the two
studies were different, even thought both versions of DA had the same amount of items.

The histogram of DA scores showed that distribution was close to normal

60

distribution. After normality test by Q-Q plot1 and Kolmogorov-Smirnov test
(Kolmogorov-Smimov = .071 , p=.20 confirmed its normal distribution), it showed that
the DA scores in this study were in a normal distribution, which was a symmetric and
bell-shaped distribution.

In addition, it was found that the correlation between the DA and the Brief CTS
was .45 (p=.000), which meant that the DA had a moderate criterion validity related to
the Brief CTS, and vise versa.

GeneraLfindingjnBrieLCISscores. As mentioned in Chapter 3, the Brief CTS
in this study included only 9 “physical aggression” items from the original CTS and the
scoring method was based on the different weighting in each violent behavior item. The
frequencies and percentage of 9 items in Brief CTS were shown at Table 11.

Even though the mean of Brief CTS was 50.12, it was found that the distribution of
the Brief CTS in samples was Poisson Distribution, which is a skewed distribution
appropriate and useful for phenomena that have a very small probability of occurring on
any particular trial, but for which an extremely large number of trials are available (SPSS,
1999b). During normality tests, the Kolmogorov-Smimov value was .223 (p<.000) and

the Q-Q Plot showed that the points did not cluster around the straight line, both of which

 

' If the sample is from a normal distribution, points would cluster around a straight line in Q-Q Plot (SPSS,

61

suggested that the Brief CTS scores in the whole samples were not in a normal

distribution, which would violate the basic assumption of lineal multiple regression while

 

 

 

 

 

 

 

 

 

 

doing the prediction.
Table 11 Frequency and Percentage in 9 Brief CTS Items
never Once twice 3-5 6-10 11-20 More (total)
times times times than 20
I. threw something at 27 8 22 30 18 6 10 121
me (22.3%) (6.6%) (18.2%) (24.8%) (14.9%) (5.0%) (8.3%) (100%)
2. pushed, grabbed, or 19 12 23 34 19 11 3 121
shoved me (15.7%) (9.9%) (19%) (28.1%) (15.7%) (9.1%) (2.5%) (100%)
3. slapped me 42 18 22 20 11 4 4 121
(34.7%) (14.9%) (18.2%) (16.5%) (9.1%) (3.3%) (3.3%) (100%)
4. kicked, bit, or hit me 26 l9 18 34 12 4 8 121
withafist (21.5%) 15.7%) (14.9%) (28.1%) (9.9%) (3.3%) (6.6%) (100%)
5. hit or tried to hit me 56 14 12 23 8 3 4 120
with something (46.3%) (11.6%) (9.9%) (19.0%) (6.6%) Q.5%) (3.3%) (100%)
6. beat me up (beat me 53 20 22 16 4 3 3 121
continuously in one (43.8%) (16.5%) (18.2%) (13.2%) (3.3%) (2.5%) (2.5%) (100%)
episode)
7. choked me 77 28 5 7 4 0 0 121
(63.6%) (23.1%) (4.1%) (5.8%) Q.3%) (0%) (0%) (100%)
8. threatened me with a 98 16 3 3 l 0 0 121
knife (81.0%) (13.2%) (2.5%) (2.5%) (0.8%) (0%) (0%) (100%)
9. usedaknife or any 104 13 2 2 0 0 0 121
killing behavior (86%) (IO-7%) (17%) (17%) (0%) (0%) (0%) (100%)

 

except choking

 

 

 

 

 

 

 

 

 

Note. There was one missing value in Item 5, so the total was 120.

Amauntnfvieleneeanddemographiedata The relationships between the

amount of violence and the demographic data would be explored as the following.

Regarding abusers’ age, after doing the Pearson correlation, it was found that the

 

1 999b)

62

 

correlation between amount of violence and abusers’ age was significantly positive but
weak (F .189, p=. 038). Therefore, it meant the older the assaulters, the larger amount of
violence toward their intimate partners, which was very different from the findings of
most other violent behaviors. The reasons might be that the nature of intimate violence is
different from other types of crime or the younger victims might not stand the intimate
violence then the older victims in Taiwan. However, all of the abusers’ education,
employment status, and ethnicity did not have significant relationship to the abusers’
amount of violence (Table 12). Moreover, it was found there was no significant
correlation between the age difference between couples and the amount of violence. All
of the above showed that other than the age of the abusers, most of the demographic data
could not significantly relate to the intimate assault, which meant there were some other
nondemographic variables more related to inmate assault than demographic variables.
Regarding the victims’ demographic data, it was found that their employment
status and ethnicity did not have significant relationship with the amount of violence,
except age of victims (1': .189, p=. 038), which meant the older victims reported
significantly more violent than the younger victims. It might be assumed that the older
victims were more conservative and had less willingness to report the intimate violence

unless the violence was too much. Since it was found that test of homogeneity of

63

variances on Brief CTS scores among levels of education reached significance, which
meant variances among different level of education were significantly heterogeneous, it

was not appropriate to do the ANOVA test (Levene = 5.465, p =.000)

Table 12 Amount of Violence and its Relationship to Demographic Data

 

 

 

 

 

 

 

 

 

 

 

 

 

 

r or F p
Abusers’ age r = .189 .038*
age difference b/w couples r = -.013 .886
abusers’ level of education F= 1.509 .182
abusers’ employment status F=38.9 .679
abusers’ ethnicity F=.914 .475
victims’ age r =.192 .036*
victims’ level of education NA NA
victims’ employment status F=1.443 .240
victims’ ethnicity F=1.l48 .340

 

Note. (1) * p<.05 ;(2) NA meant not available, because it was not homogeneous on

variances.

Di: ‘r‘ o, It.» '11 e 1“: a ._-. l ‘_ .1..-“ I..- -51.. Orr. ‘

Since the Danger Assessment scale was originally developed in the America and
was designed for assessing the lethality of intimate abuse, comparing its use in Taiwan
and in America would be focus the comparison of abusers with and without lethal
violence in both areas. It was found that Item 2 (increased severity of violence), Item 3
(choking), Item 6(l)(he threatened to kill), and Item 12(he threatened to commit suicide),

were significantly different between the male intimate abusers with and without lethal

64

violence among the Taiwan samples (Table 13). Therefore, all of these four items could

be good predictors of lethal violence in Taiwan.

Table 13 Mean Differences between Abusers with and without Lethal Violence in 15
DA Items among Taiwan Sample by T-test

 

Taiwan Sample

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Abusers with Abusers without p
lethal violence lethal violence in t-test
08:50) (N=71)
1. increased frequency of .56 .44 .184
violence
2. increased severigl of violence .65 .46 .035*
3. choking ___ .78 .17 .000***
4. forced sex .52 .46 .554
5. drug abuse .02 .0845 .099
6(1). he threatened to kill .70 .51 .032*
6(2). she believes to be killed .56 .42 .138
7. alcohol abuse .44 .34 .259
8. control daily life .56 .55 .908
9. assault while prtgnant .41 , .28 .170
10. violent jealousy .56 .41 .098
11. she threatened/tried to .52 .45 .457
commit suicide
12. he threatened/tried to commit .36 .15 .013*
suicide
13. violent to child .42 .41 .951
14 violent outside .33 .27 .489

 

Note. ***p<.001, **p<.01, *p<.05

C I O O .I .0 O
,‘0 0'60t0-_10 0'1 ‘1». - Del. 01"“ o

BredictedamountefxielerreeiahdeffllthDA As mentioned earlier, the

CTS scores were in poisson distribution. For predicting poisson distribution scores, it

should not use the regular regression model i.e. ordinary least square regression, which

were based on normal distribution. Instead, only non-parametric regression techniques
can be used, such as TOBIT or Poisson Regression. TOBIT was suggested by Straus &
Gelles (1990, p453) as one technique to do the regression on the severity weighted scale
as in the Brief CTS, but TOBIT was noted to be used only in continuous variables
(Maddala, 1983; Long, 1997). Poisson regression is the regression model using
maximum-likelihood estimation (MLE) to predict the count data in the poisson
distribution. The author was convinced that the weighted amount of violent behavior
should be still a discrete variable. Therefore, in this study, the author would use Poisson
Regression for predicting the amount of violence (Brief CTS) by the Danger Assessment
scale.

After predicting the amount of violence (Brief CTS) by 15 items in the Danger
Assessment scale by poisson regression, it was found that Item 5, Item 10, and Item 12
did not reach the significance in the prediction (p values were .76, .58, and .39,
respectively), and the R-square was .2965, which meant the 15 DA items could explain
29.65% of variance in the amount of violence by the Brief CTS (Table 14). Meanwhile,
after doing the forward stepwise regression analyses on 15 DA items to Predict the
Amount of Violence in Brief CTS, it was confirmed that the above three items should be

excluded, because their predictions did not reach significant level (p values

66

were .125, .364, and .478, respectively)(Table 15). After excluding those three items, the
other 12 items in the DA could was found the R-square was .2916, which meant that the
12 items in the DA could explain 29.16% of variance of the amount of violence by the
Brief CTS (Table 16). However, it was found that 4 negative regression coefficient
existed, which might not be reasonable. There might be several reasons for this situation.
First, the number of samples in this study might not be large enough for the 15
independent variables in the DA to run the regression. Normally, it is better to add 20
subjects while add each independent variable (Wei Pan, personal communication, March
8, 2000). Second, the collinearity among these independent variables might exist and it is
difficult to explain some unreasonable phenomenon, such as creating negative regression
coefficients or/and causing some supposedly important predictors not to reach significant
level in their responded regression coefficient (Cohen, 1983, p115).

Several methods were suggested to solve the collinearity in Chen (1997, p8-12),
such as reducing the number of independent variables, using ridge regression, and using
principle component analysis and then using these principle components as independent
variables to predict the outcome variable. Of these solutions, the author would choose to
use the factor analysis to find the principle components among these 12 DA items. Five

principle components existed (Table 17) and each of them reached the significant level in

67

prediction toward the amount of violence and regression coefficients were all positive,
which meant the more tendency on the components the more the amount of violence
presented (Table 18). Among those five components, five DA items (threaten to kill,
belief to be killed, violent outside, alcohol abuse, and forced sex) were grouped in
Component 1, two DA items (increase severity of violence and increase frequency of
violence) were grouped in Component 2, one DA item (choking) was in Component 3,
two DA items (violence to child and assault when pregnant) were grouped in Component
4, two DA items (she wanted suicide and he wanted suicide) were grouped in Component
5. The R-square in this regression was .2686, which meant those 5 components could
explain 26.86% of variance in the amount of violence.

Furthermore, it was found that the correlation between the scores of the DA and

Brief CTS reached significant level (r = .432, p = .000).

68

Table 14 Predicting the Amount of Violence (Brief CTS) by 15 Items in Danger

Assessment by Poisson Regression

 

 

Cts Coef. Std. Err. z P>|zl [95% Conf. Interval]
da,1 increase frequency .57 .03 18.60 0.00 .51 .63
da_2 increase severity -.10 .03 -3.18 0.00 -.17 -.04
da_3 choking .48 .03 16.14 0.00 .42 .54
da_4 forced sex -. 10 .03 -3.08 0.00 -.16 -.04
da_5 drug abuse .019 .06 0.30 0.76 -.11 .14
da_6_l threat to kill .28 .04 6.81 0.00 .20 .36
da_6_2 believe to be killed .23 .04 6.57 0.00 .16 .30
da_7 alcohol abuse .34 .03 11.12 0.00 .28 .40
da_8 control daily life .02 .03 0.55 0.58 -.05 .08
da_9 assault when pregnant -.00 .00 -4.83 0.00 -.00 -.00
da_10 violent jealousy -.03 .03 -0.94 0.35 -.09 .03
da.ll she want suicide .18 .03 6.28 0.00 .12 .23
da_l2 he want suicide -.078 .03 -2.38 0.02 -.14 -.014
da_l3 violent to child .22 .03 7.87 0.00 .17 .28
da_14 violence outside .20 .04 5.95 0.00 .14 .27
_cons 2.79 04 70.63 0.00 2.71 2.87

 

Note. Number of obs = 116 (since missing values were excluded); LR ch12(15) = 1877.57;

Prob > chi2 :- 0.0000; Log likelihood : -2227.0846; Pseudo R-square := 0.2965

Table 15 Forward Stepwise Regression Analyses on 15 DA Items to Predict the

Amount of Violence in Brief CTS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

R-square Stepwise b-value p Suggest
change in to add
R-square or
Exclude
DAl increase frequency of violence .0631 .520 .000 Add
DA2 increase severity of violence .0664 .0051 -.090 .003 Add
DA3 choking_ .1622 .0958 .646 .000 Add
DA4 forced sex .1664 .0042 .141 .000 Add
DA5 drug abuse .1668 .0004 -.094 .125 Exclude
DA6(1) threatened to kill .2260 .0592 .593 .000 Add
DA6(2) believed to be killed .2410 .0150 .317 .000 Add
DA7 alcohol abuse .2719 .0309 .398 .000 Add
DA8 control daily life .2720 .0001 -.025 .364 Exclude
DA9 assault when pregnant .2736 .0016 -.003 .003 Add
DAIO violent jealousy .2768 .0032 -.023 .478 Exclude
DAll she wanted suicide .2822 .0054 .016 .000 Add
DA12 he wanted suicide .2834 .0012 -.088 .006 Add
DA13 violence to child .2910 .0076 .199 .000 Add
DA14 violent outside .2965 .0055 .200 .000 Add

 

 

69

 

Table 16 Predicting the Amount of Violence (Brief CTS) by 12 Items in Danger
Assessment scale by Poisson Regression (excluded Item 5, 8, and 10,)

 

 

Cts Coef. Std. Err. z P>lzl [95% Conf. Interval]

da_l increase frequency .58 .03 19.19 0.00 .52 .64
da_2 increase severity —.08 .03 -2.63 0.01 -.14 -.02
da_3 choking .50 .03 17.48 0.00 .44 .55
da_4 forced sex -.07 .03 -2.41 0.02 -.13 -.01
da_6_1 threat to kill .28 .04 6.94 0.00 .20 .36
da_6_2 believe to be killed .21 .03 6.12 0.00 .14 .28
da__7 alcohol abuse .32 .03 10.39 0.00 .26 .38
da_9 assault when pregnant -.01 .00 487 0.00 -.01 -.Ol
da_ll she want suicide .19 .03 6.88 0.00 .14 .25
da_12 he want suicide -.07 .03 2.39 0.02 -.13 -.01
da~13 violent to child .20 .03 7.33 0.00 .15 .25
da_14 violence outside .18 .03 5.57 0.00 .12 .25

_cons 2.76 04 72.63 0.00 2.69 2.84

 

Note. Number of obs = 118(since missing values were excluded); LR chi2(l2) : 1881.94;
Prob > chi2 = 0.0000; Log likelihood 2 -2286.164; Pseudo R—square = 0.2916.

Table 17 Five Principle Components among the 12 Items in Danger Assessment scale

Component Matrix ‘

 

 

 

 

Component
1 2 3 4 5

da6(1)threat to kill .704 .227 - 419
da14_violent outside .652 - 211 -.292 -.320
da6(2)belief be killed .639 .341 -.407 .247 .150
da7_alcohol abuse .548 -.479 .212 -.l60
da4_force sex .453 -.291 .280 -.387 - 319
da2 increase sever or
not‘ .162 .765 .353
dal increase fre or
not— q .147 .627 .396 -.149 -.439
da3_choke .343 .422 .185
dal3_violent to child .270 .245 .674 .159
da9 assault when
pregnant .255 -.423 .384 .455
dall_she try suicide .351 .111 .121 -.388 .598
dal2_he wanted suicide .333 .104 -.382 .416

 

 

 

 

 

 

 

Extraction Method: Principal Component Analysis by varimax rotation.
a. 5 components extracted.

70

Table 18 Predicting the Amount of Violence (Brief CTS) by 5 principle components
among the 12 Items in Danger Assessment scale by Poisson Regression

 

 

Cts Coef. Std. Err. z P>lz| [95% Conf. Interval]
Componentl .28 .01 19.83 0.00 .25 .30
Component2 .27 .01 21.13 0.00 .25 .30
Component3 .24 .01 17.52 0.00 .21 .26
Component4 .29 .01 23.27 0.00 .27 .32
Component5 . 15 .01 11.68 0.00 . 12 .17

_cons 3.79 .015 253.02 0.00 3.76 3.82

 

Note. Number of obs = 114(since missing values were excluded); LR ch12(5) = 1707.76;
Prob > ch12 = 0.0000; Log likelihood : 23253984; Pseudo R-SQUARE = 0.2686.

LethaLYieleneedetaAnalyms
Lethalyiolenceandareas The lethal violence was directly recoded from the Brief

CTS’s Item 7 (choking) and Item 9 (used a knife or any killing behavior except choking,
which was extended from using a knife or firing a gun in original CTS). During the cross
analysis between lethal violence and areas, the total subjects were split into three areas,
which were Kaohsiung City, F onshan Township, and the rural area in Kaohsiung County.
Fonshan was single out, because it is a very prosperous township adjacent to Kaohsiung
City and it is much more like a city area than rural area even thought it is in Kaohsiung
County. It was found that the appearance of lethal violence were significantly different
within the city, suburban town, and rural areas as shown in Table 19 (chi-square = 7.058,
p=.029). There might be two possible reasons. First, it is more conservative in the rural

areas, since the victims might be more willing to report domestic violence only when it

71

was very severe to them. Second, rural areas rrright not have the resource to report their

victimization or to learn about the recent implementation of new Domestic Violence

Prevention Law.

Table 19 area * lethal violence or not Crosstabulation

area * lethal violence or not Crosstabulation

 

 

 

 

 

 

 

 

 

 

 

lethal violence or
not
no ,yes Total

area K cify Count 54 28 82
% within area 65.9% 34.1% 100.0%

K county_fonshan Count 8 6 14

% within area 57.1% 42.9% 100.0%

K county_rural area Count 9 16 25

% within area 36.0% 64.0% 100.0%

Total Count 71 50 121
% within area 58.7% 41.3% 100.0%

 

Note, Chi-square = 7.058, p= .029*

Lethalxieleneeanddemographiedata The relationships between the times of

lethal violence and the demographic data would be explored as the following.

It was found that all of the abusers’ age, level of education, employment, and
ethnicity could not have significant relationship with the times of lethal violence (Table
20). Moreover, the age difference between the couples did not have significant correlation
to the times of lethal violence.

Regarding the victims’ demographic data, it was found all of the victims’ age, level

of education, employment status, or ethnicity did not have a significant relationship with

72

 

the times of lethal violence (Table 20).

Table 20 Time of Lethal Violence and its Relationship to Demographic Data

 

 

 

 

 

 

 

 

 

 

 

 

 

 

r or F p

abusers’ age r = -.016 .861

e difference b/w couples r = -.067 .473
abusers’ level of education F = .738 .620
abusers’ emplgyment status F =.275 .760
abusers’ ethnicity F=.32l .900
victims’ age r =.007 .937
victims’ level of education F = .458 .838
victims’ employment status F=1.229 .296
victims’ ethnicity F=.491 .782

MW Even though the lethal violence was

recoded from the Brief C TS, it was still worthy to explore the mean differences of the
scores on the DA and Brief CTS between the intimate abuses with and without lethal
violence. It was found that e mean differences of the scores on the DA and Brief CTS
between the intimate abuses with and without lethal violence were significant (t = 3.598,
p = .001; t= 3.826, p = .000, respectively) (Table 21). However, it had to be noted that the
range of both groups in scores of the DA and Brief CTS were still large. It was showed in
Table that the highest Brief CTS score in abuses without lethal violence was 207,
which meant this abuser might only use high amount of minor violence (Item 1 through 6,

and Item 8) but not for any lethal violence.

73

 

Table 21 Mean Differences of Brief CTS and DA between Abusers with and without
lethal violence

 

 

 

 

 

(number of samples) Brief CTS DA
Abuser with lethal 50 72.42 1. 34
violence 413449) (1 ~l 3)
Abuser without lethal 71 34.42 4.38
violence (0~207) (0~12)
(total = 121) t=3.598 t = 3.826
p= .001 ** p =.000***

 

 

 

 

WW Two measures

were used to describe the predictive accuracy of the risk scale: (a) r, the correlation
coefiicient, and (b) the area under the receiver operating characteristic (ROC) curve
(Hanson, 1997). Receiver operating characteristic (ROC) was first developed in
communication technology and in signal detection theory in psychophysics. Not until the
mid-90$, was the ROC applied in predicting violence (Quinsey, Harris, Rice, & Corrnier,
1998). The ROC curve is a useful way to evaluate the performance of classification
schemes in which there is one variable with two categories by which subjects are
classified. The area under the ROC curve was used as the primary measure of predictive
accuracy (Mossman, 1994; Rice & Harris, 1995). The ROC curve plots the hits
(accurately identified recidivists) and false alarms at each level of the risk scale. The
area under the ROC curve can range from .50 to 1.0, with 1.0 indicating perfect

prediction (no overlap between recidivists and non-recidivists) and .5 0 indicating

74

prediction no better than chance (Figure 2; note: The diagonal of this figure is .50, which
means guess by chance). In general, the ROC area can be interpreted as the probability
that a randomly selected recidivist would have a more deviant score than a randomly
selected nonrecidivist. The ROC area has advantages over other commonly-used
measures of predictive accuracy (e. g., percent agreement, correlation coefficients) since it

is not constrained by base rates or selection ratios (Hanson & Thornton, 1999).

Figure 2 Receiver Operative Characteristic (ROC)

 

1

0.8 -
Hit
Rate

 

0.6

0.4

0.2 ’

 

 

O 0.2 0.4 0.8 O 8 1

False Alarm Rate
Note. (1) Hit rate is true positive of the prediction, and false alarm is false positive of
prediction. (2) From Quinsey, V. L., Harris, G. T., Rice, M. E., & Cormier, C. A. (1998).
Violent oflenders: Appraising and managing risk. Washington, DC: American
Psychological Association.

75

In this study, whether or not the lethal violence presented was reported by the
victims would be predicted by the DA scale through the predictive accuracy of the ROC
curve. It was found that the area under the ROC curve of the DA was .753, which was
moderate and satisfactory accuracy, and the area under the ROC curve of the Brief CTS
was .718, which was a lower moderate and still a satisfactory accuracy (see Figure 3 and
Table 22), even though the lethal violence was directly recoded from the Brief CTS’s
Item 7 (choking) and It: m 9 (used a knife or any killing behavior except choking). The
reason for this situation would be that the weighing system of physical violence in the
CTS could not clearly differentiate the lethal violence and non-lethal violence, even

though the more severe violence was weighted higher.

wr‘

Figure 3 Predicting Lethal Violence by DA and CTS in ROC Curve Prediction Accuracy

ROC Curve

 

1.00

.75‘

.50'
Source of the Curve

 

ei”l ° Reference Line
.25‘ ,-' . ..

 

DA

0.00 E . ' . ” Cts score
0.00 .25 .50 .75 1.00

 

 

Sensitivity

 

l - Specificity
Diagonal segments are produced by ties.

76

Table 22 Area Under the ROC Curves for Predicting Lethal Violence by CTS and DA

Area Under the Curve

 

 

 

Asymptotic 95%
Test Result Asymptotic Confidence Interval
Variable(s) Area Std. Errord Sig.3 Lower Bound Upper Bound
cts score .753 .043 .000 .668 .838
DA .718 .047 4000 .627 .810

 

 

 

 

 

 

 

 

The test result variable(s): cts score , DA has at least one tie between the
posrtrve actual state group and the negative actual state group. Statistics may
be biased.

a. Under the nonparametric assumption
b. Null hypothesis: true area = 0.5

"
i.
O

"q' '13 ‘ J vioer ' n D‘ e 9 .31 01' gist 0‘ "1
It was found that the correlation coefficient between lethal violence and DA was .364

(p<.000), and the correlation coefficient between lethal violence and Brief CTS was .333

(p<.000) (Table 23).

Table 23 Bivariate Correlation within CTS, DA, Lethal Violence

 

 

 

 

 

 

 

 

Correlations
lethal
violence
cts score DA or not
cts score Pearson Correlation 1.000 .450*‘ .333*
Sig. (2-tailed) . .000 .000
N 121 121 121
DA Pearson Correlation .450*‘ 1.000 .364*
Sig. (2-tailed) .000 . .000
V 121 121 121
lethal violence or not Pearson Correlation .333*‘ .364*i 1.000
Sig. (2-tailcd) .000 .000 .
N 121 121 121

 

 

**. Correlation is significant at the 0.01 level (2-tailed).

PredictingdefhalJieleneehxflrteffsanBneLClS. It would be worthy to

categorize the abusers based on the amount of violence for the purpose of helping social

77

 

workers identify the lethality risk to domestic violence victims. The author created three

categories as shown in Table 24. The cutoffs were selected as 25%, 66%, 50%, and 75%

of the Brief CTS scores, which were 15, 32, 43, and 60, respectively (Table 25).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 24 Three Categories for Stratifying the CTS Scores
Three Categorizing Level of amount of violence Cutoff in CTS % of Chi-
Ways lethal square
violence

CTS_66 low violent amount abuser Lowest —42 27.5% 18.60***
(cutoff at 66%) high violent amount abuser 43-highest 68.3%
CTS_CU3 low violent amount abuser Lowest-14 12.9% 17.39***
(cutoff at 25% & 75%) mid violent amount abuser 15-59 44.1%

high violent amount abuser 60-highest 64.5%
CTS_CU3A low violent amount abuser Lowest-31 24.2% 16. l 1 ***
(cutoff at 50% & 75%) mid violent amount abuser 32-59 53.6%

high violent amount abuser 60-highest 64.5%
Note. ***p< .001

Table 25 Descriptions in Scores of Brief CTS and DA

Statistics

 

 

 

 

cts score DA
N Valid' 121 121
Missing 0 0
Mean 50.12 6.04
Median 32.00 6.00
Mode 03 7
Std. Deviation 56.41 3.01
Percentiles 25 15.00 4.00
50 32.00 6.00
66 43.52 7.00
75 60.00 8.00

 

 

 

a.

Multiple modes exist. The smallest value is

shown

78

 

Figure 4 Predicting Lethal Violence by three Categories of CTS in ROC Curve
Prediction Accuracy '

 

 

 

 

 

 

ROC Curve
1.00 y - (I;
.75‘
Source of the Curve
” Reference Line
.50'
° CUT_CU3A
f? ....
E .25; ° cts score
3 CUT_QB
S: .---
B 0.00, _ . - ° CUT_CU66
0.00 .25 .50 .7" 1.00

1 - Specificity

Diagonal segments are produced by ties.

Table 26 Areas under ROC Curves for three Categories in CTS Scores to Predict the
Appearance of Lethal Violence

Area Under the Curve

 

 

 

 

 

Test RCSUII’. Area

CULCU66 .688
CUT_CU3 .703
cts score .753
CUT_CUBA .694

 

The test result variable(s): CUT_CU66, CUT_CU3.
cts score . CUT_CU3A has at least one tie between
the positive actual state group and the negative
actual state group. Statistics may be biased.

After comparing the three categories’ predicting the appearance of lethal violence,
the CUT_CU3 category, which cutoffs were at 25% and 75%, had the largest ROC area

with .703 (Table 26 and Figure 4). Therefore, this way would be recommended for future

79

 

use in Taiwan.

PredictinglethaLyialeneehuutafisinDA. The author created three categories as

shown in Table 27. The cutofis were selected as 25%, 66%, 50%, and 75% of the Danger

Assessment scale which were 4, 6, 7, and 8, respectively (see Table 25).

Table 27 Three Categories for Stratifying CTS Scores

 

 

 

 

 

 

 

 

 

 

 

 

 

Three Categories Level of amount of violence Cutoff in DA % of Chi-
lethal square
violence

DA_66 Low violent amount abuser Lowest —6 25.4% 15.75***

(cutoff at 66%) High violent amount abuser 7-highest 61.1%

DA_CU3 Low violent amount abuser Lowest-3 20.0% 10.742”

(cutoff at 25% & 75%) Mid violent amount abuser 4-7 37.9%

High violent amount abuser 8-highest 60.5%
DA_CU3A low violent amount abuser Lowest-5 21.2% 15.86***
(cutoff at 50% & 75%) mid violent amount abuser 6-7 51.6%

high violent amount abuser 8-highest 60.5%

 

Note. **p< .01, ***p< .001

80

 

Figure 5 Predicting Lethal Violence by three Categories of DA Scores in ROC Curve

Prediction Accuracy

 

 

 

 

ROC Curve
1.00 -
.754 ‘4’
Source of the Curve
’f/l ‘ Reference Line
.50‘ 1' .___.
° DA_CU3A
2: ,' -----
.i .25- MBA-w
.2 DA_66
e r -
v: 0.00 ' . fi fi ° DA
0.00 .25 .50 .75 1.00

1 - Specificity

Diagonal segments are produced by ties.

Table 28 Areas under ROC Curves for three Categories in DA Scores

Area Under the Curve

 

 

 

 

Testhesult Area

DA .714
DA_66 .682
DA_CU3 .661
DA_CU3A .694

 

 

The test result variable(s): DA, DA_66. DA_CU3.
DA_CU3A has at least one tie between the
positive actual state group and the negative
actual state group. Statistics may be biased.

After comparing the three categories’ predicting the appearance of lethal violence,
the DA_CU3A way, which cutoffs were at 50% and 75%, had the largest ROC area

with .694 (Table 28 and Figure 5). Therefore, this way would be recommended for future

81

use in Taiwan.

PredicnngdethalltolenceerefinedDA. It would be worthy to examine the

DA’s prediction toward lethal violence by logistic regression. The R-square, which is a
coefficient of determination, was found to be .3983, which meant the DA could explain
39.83% of the variance in the appearance of lethal violence. However, it was also found
that only Item 3 and Item 12 in the revised and translated DA reached the significant
level of .05. There might be several reasons for this situation. First, the number of
samples in this study might not large enough for the 15 independent variables in the DA
to run the regression. Normally, it is better to add 20 subjects while add each independent
variable (Wei Pan, personal communication, March 8, 2000). Second, the collinearity
among these independent variables might exist and it is difficult to explain some
unreasonable phenomenon, such as creating negative regression coefficients or/and
causing some supposedly important predictors not to reach significant level in their
responded regression coefficient (Cohen, 1983, p115) Regarding the collinearity, it was
confirmed in that , of a total of 105 bivariate correlation among 15 DA items, 20 bivariate
correlation reached significant level (Table 29). Third, both dependent and independent

variables were dummy variables, which cause more problems in accuracy of prediction

than continuous variables would do.

82

Several methods were suggested to solve the collinearity in Chen (1997, p8-l2),
such as reducing the number of independent variables, using ridge regression, and using
principle component analysis and then using these principle components as independent
variables. Of these solutions, the author would choose to reduce the number of
independent variables. After comparing several ways of reducing the number of
independent variables, four items (Item 1, 3, 6(1), and 12. See Table 30. It was noted that
Item 4 was excluded, since it could not reach significance between abusers with and
without lethal violence as previously mentioned) in the DA remained as the independent
variables to run the logistic regression, which was found the R-square was .3523, which
meant they could explain 35.23% of the variance of the appearance of lethal violence
(Table 31). Based on this finding, a Brief DA with these four items (Item 1, 3, 6(1), 12)
could be formed. These four items were chosen because their significant levels were
relatively small (the cutoff for screening the DA items was .20, since sample size in this
study was not big enough).

After doing the ROC accuracy test on a new variable composed by these four items,
the area under the curve .of this composed variable was .801, which was still smaller than
the area under the curve of Item 3 (choking behavior) (.805), even thought choking

behavior is not the only category in lethal violence.

83

 

8V...» .ch... daz

 

2:80 856$ 31%

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

no. 2% 9 6203 21%
m 8. 8.- oEosm :83 on 21%
8. 8. ......om. oEoSm :83 new :18
......em. 8. m _. 3. 3228.4 E20; 21%
2. ...om. no. 8.- :. Eds-65 sons :33“ 81%
:2. 8. .2. 8.- :3. 3. a: ran 65:8 miss
:3. 2. S. 2. ...mm. :8. 8. 33m 68% >18
.39”. .3. Q. 8. .2. 8.- 2. 2. 8%. on 9 o>ozoemuoumn
.38. S. 3.. 2. .... _ m. 8.- 8. 2. Ice. :3 8 8857018
8. La. 8. 8.- 8. 8.- S. 2. 2.- 8.- 63% wee miss
:8. mo. 8. 2. :8. 8. :3. 38m. 8. 2. 8. x8. BBQ v18
8. mo. 8. 8. S. E. 8.- 2. 8. E. 8.- E. 8205 Minn
8.- :. 8. 8. 8.- 8.- 2. 2.- S. S. m ..- 8.- 8. 5.298 98.65 mine
8.- 8.- 8. 8. 8. 3.- E. 8.- 8. 8. 2.- 2. 8.- tan. 35:38 888E ~18
6350 EEO 868m @223 8835 8:9 >
8 9 ES: Ema $228.4 :23 fl: 8% 83m on o. :3 83m x8 m >Eo>om 25:38
:20; 8203 on new :86: 83% 638 6:8? 26:3 9 28:: $8 388 :29? 8885 8885
318 21% $18 :18 218 81% wig mien mic-.8 181% misc v18 mtg ~18 ~18

 

was: <0 2 ”:25 22.22.80 oEBZm .8 M Samoa am 298.

 

84

Table 30 Predicting the Appearance of Lethal Violence by 15 Items in Danger Assessment
by Logistic Regression ‘

 

 

 

Lethal violence Odds Ratio Std. Err. z ‘ P>lz| [95% Conf. Interval]
da_l increase frequency 2.38 1.48 1.40 0.16 .71 8.03
da_2 increase severity 1.63 1.04 0.77 0.44 .47 5.66
da_3 Choking 33.25 22.94 5.08 0.00 8.60 128.54
da_4 forced sex . .45 .27 -1.32 0.19 .14 1.47
da_5 drug abuse .21 .36 0.91 0.36 .01 6.00
da_6_1threat to kill 1.93 1.50 0.85 0.40 .42 8.81
da_6_2believe to be 121116690 .65 —0.14 0.89 .22 3.71
da_7 alcohol abuse 1.61 .98 0.79 0.43 .49 5.30
da_8 control daily life 1.27 .87 0.35 0.73 .33 4.85
da_9 assault when pregnant .99 .02 -0.65 0.52 .95 1.02
da_10 violent jealousy 1.44 .93 0.56 0.58 .40 5.10
da_ll she want suicide .73 .41 -0.57 0.57 .24 2.18
da_12 he want suicide 7.03 5.04 2.72 0.01 1.72 28.67
da_l3 violent to child .73 .41 —0.56 0.57 .25 2.17
da_14 violence outside .87 59 020 0.84 .23 3.32

 

 

Note. LR 0111205) = 62.05, Prob > ch12 = 0.0000, Log likelihood = 46.86906, Pseudo R-
square = 0.3983

Table 31 Predicting the Appearance of Lethal Violence by 4 Items in Danger
Assessment by Logistic Regression

 

 

 

Lethal violence Odds Ratio Std. Err. z P>lz| [95% Conf. Interval]
da_l increase frequency 2.55 1.32 1.81 0.07 .92 7.03

da_3 choking 21.53 11.62 5.69 0.00 7.48 61.99

da_6_1 threat to kill 1.67 .85 1.02 0.31 .62 4.50

da_12 he want suicide 4.45 2.66 2.50 0.01 1.38 14.36

 

 

Note; Number of obs = 121, LR chi2(4) = 57.81, Prob > ch12 = 0.0000, Log likelihood = -
53.133535, Pseudo R-square = 0.3523.

85

Figure 6 ROC Curves for Predicting Lethal Violence by Item 1, 3, 6(1), & 12 in DA

ROC Curve

Source of the Curve

Reference Line

BRIEF_DA

” da12_he wanted suici

 

1.00

 

° da6(1)threat to kill

. da3_choke

——_—_.

“ dal_increase freq or

 

not

 

Sensitivity

 

0.00 .25 .50 .75 1.00

1 - Specificity

Diagonal segments are produced by ties.

Table 32 Area under ROC Curves for Predicting Lethal Violence by Item 1, 3, 6(1), &
12 in DA

Area Under the Curve

 

 

 

 

est Result Area
dal increase fre or
not— q .562
da3_choke .805
da6(1)threat to kill .596
da12_he wanted suicide .603
BRIEF_DA .801

 

 

The test result variable(s): da1_increase freq or
not, da3_choke, da6(l)threat to kill, da12_he wanted
suicide, BRIEF_DA has at least one tie between the
positive actual state group and the negative actual
state group. Statistics may be biased.

For the purpose of finding the difference between logistic regression and poisson

regression, the author recoded the times of presenting lethal violence from the Brief CTS

86

(Item 7 “choking” and Item 9 “using knife toward you andany other killing behavior”
were included. Item 8 “threatened you with a knife” was excluded, it did not reach killing
behavior), in addition to the appearance of lethal violence. It was found, however, after
doing the possion regression to predict the times of lethal violence by 15 DA items, the
R-square lowered to .1973 (Table 33), while the R-square was .3983 after doing the
logistic regression to predict the appearance of lethal violence by the same 15 DA items.
The reason for this situation might result from the three same reasons as previously
pointed out-- too small sample size, the problem of collinearity or the problem of dummy
variables. Moreover, the times of lethal violence was found as a skewed distribution
(Figure 7), which might also cause poisson regression lower its R-square, compared the

use of logistic regression, which could balance the skewed distribution.

87

Table 33

by Poisson Regression

Predicting the Times of Lethal Violence by 15 Items in Danger Assessment

 

 

 

 

Lethal violence Odds Ratio Std. Err. z P>|z| [95% Conf. Interval]
da_l increase frequency .29 .22. 1.33 0.19 -.14 .71
da_2 increase severity -.16 .23 0.70 0.49 -.61 .30
da_3 choking 1.15 .23 5.05 0.00 .70 1.60
da_4 forced sex -.35 .23 -1.54 0.12 -.80 .09
da_5 drug abuse -1.29 .74 -1.68 0.09 -2.69 .21
da_6_1 threat to kill .75 .31 2.46 0.01 .153 1.36
da_6_2 believe to be killed .22 .24 0.89 0.4 -.25 .70
da_7 alcohol abuse .35 .22 1.56 0.11 -.08 .79
da_8 control daily life .18 .25 0.73 0.47 -.31 .67
da_9 assault when pregnant -.016 .012 -1.37 0.17 ~.04 .01
da_10 violent jealousy .30 .23 1.29 0.20 -.16 .76
da_ll she want suicide .24 .21 1.15 0.25 -.17 .64
da_12 he want suicide -.10 .23 0.42 0.67 -.56 .36
da_13 violent to child -.17 .20 -0.87 0.38 -.57 .22
da_13 violent to child -.06 .25 -0.24 0.81 -.54 .426
_cons -1.60 .32 -5.03 0.00 -2.22 -.98
_N_o_te, Number of obs = 116; LR chi2(15) = 80.48; Prob > chi2 = 0.0000 ; Log likelihood = -

163.71767; Pseudo R-SQUARE = 0.1973.

Figure 7 Histogram of Times of Lethal Violence in Taiwan Samples

120‘

60

40

20

 

 

 

2.5 5.0

7.5

time of lethal violence

10.0

12.5

88

Mean = 1.0
[N = 121.00

 

Std. Dev = 1.91

 

5.:11 Il'll'E .15]
Lmemrcentagmflbemgabnsedafieunhabmng. It was found that the mean of

the time percentage of being abused afiercohabiting was .79 years with the minimum
of .01 years and maximum of 20 years. Moreover, it would be worthy to explore whether
or not the time percentage of being abused afier cohabitation could predict the amount of
violence and the appearance of lethal violence. (a) Whether or not the time percentage of
being abused after cohabitation could predict the amount of violence: It was found that
the correlation between the two variables was not significant (1': .003, p = .975), poisson
regression coefficient was not significant (b= .003, p= .690, R-square :00). (b) Whether
or not the time percentage of being abused after cohabitation could predict the appearance
of lethal violence: It was found that the correlation between the two variables was not
significant (t= .917, p = .361), logistic regression coefficient was not significant (b= -.221,
p=.537, R-square =.01), and the area under the ROC curve was .469 (which meant this
prediction was worse than guess by chance) (see Table 38).

u ’- 1.00‘000 1.4 err-g a..- '0'1‘11.‘ Onascaleof
0 to 10, the mean of victims’ self assessment on the chance of being re-abused in the
fiiture was 6.17 with the responses in the minimum of 0 and maximum of 10, which was

not a normal distribution (Kolmogorov-Smimov= .167, p=.000). Moreover, it would be

89

worthy to explore whether or not the amount of violence and the appearance of lethal
violence could predict victims’ self assessment on the chance of being re-abused. (a)
Whether or not victims’ self assessment on the chance of being re-abused could predict
the amount of violencezi It was found that the correlation between the two variables was
significant (F .267, p = .003), poisson regression coefiicient was significant (b= .002
(exp[.002]= 1.002), p=.000, R-square =.028). (b) Whether victims’ self assessment on the
chance of being abused could predict the appearance of lethal violence: It was found that
the correlation between the two variables was not significant (t= 1.718, p = .088), but
logistic regression coefficient was significant (b= 1.148, p= .031, R-square= .030).
Therefore, the above results showed that both the amount of violence and the appearance
of lethal violence could significantly predict the victim’s self assessment on the chance of
being re-abused (Table 38).

AWE; While asked whether or not the victims were stalked
by the abusers in the last one-year period, 39 (32.2%) victims responded they had been
Stalked in the last year, and 52 (43%) responded none, while 28 (23.1%) responded they
did not know (Table 34). It was worthy to explore if times of being stalked could predict
the amount of violence and the times of appearing lethal violence. (a) Whether or not the

times of being stalked could predict the amount of violence: It was found that the

90

correlation between the two variables was not significant (F -.016, p = .879), poisson
regression coefficient was not significant (b= -.027 (exp[—.027]= .97), p=.106, R—square
=.001). (b) Whether or not being stalked could predict the appearance of lethal violence:
It was found that the correlation between the two variables was significant (Chi-
square=2.052 p= .358) and the logistic regression coefficient was not significant (b= 1.36,
p= .208). However, while predicting the times of lethal violence by poisson regression, it
was found that the poisson regression coefficient was significant (b= .387 (exp[.387]=
1.42), p=.001, R-square =.034), and the area under the ROC curve was .565 (which

meant this prediction accuracy was low) (Table 38).

Table 34 The Victims’ Responses on Whether Been Stalked or Not

stalking or not

 

 

 

 

 

 

 

 

 

Valid Cumulative
Frequency Percent Percent Percent

iVElid no in last yr 52 43.0 43.7 43.7

1~2 last yr 15 12.4 12.6 56.3

at least 3 last yr 24 19.8 20.2 76.5

don't know 28 23.1 23.5 100.0

Total 119 98.3 100.0
Missing System 2 1.7
Total 121 100.0

mmmmhmmmmmmn Whether or not the

abusers threatened to hurt the victims’ family of origin and the kids to prevent the

victims’ leaving them was suggested by a medical doctor in Kaohsiung City. Chih-chung

91

Huang. In his clinical practice for the female victims, he found these two items might be
related to the amount of violence and/or the appearance of lethal violence and deserves to
be explored (personal communication, August 17, 2000). It was found that 45 (37.2%)
victims responded their abusive partners did threaten to hurt their family of origin to
prevent the victims’ leaving them, while 75 (63.0%) responded not (Table 35). It was
worthy to explore threatening to hurt victim’s family of origin could predict the amount
of violence and the time of appearing lethal violence. (a) Whether threatening to hurt
victim’s family of origin could predict the amount of violence: It was found that the
correlation between the two variables was not significant (t= .182, p = .856), poisson
regression coefficient was not significant (b= -.408 (exp[-.408]= .66), p=.127). (b)
Whether or not threatening to hurt victim’s family of origin could predict the appearance
of lethal violence: It was found that the correlation between the two variables was not
significant (Chi-square=.057, p= .811), logistic regression coefficient was not significant

(b= 1.096 , p=.811).

92

Table 35 Whether or Not the Abusers Threatened to Hurt Victims’ Family of Origin

 

 

 

 

 

 

 

 

 

HURPALW
Valid Cumulative
Frequency Percent Percent Percent ‘

Valid no 75 62.0 62.5 62.5

yes 45 37.2 37.5 100.0

Total 120 99.2 100.0
Missing System 1 .8
Total 121 100.0

Abusers-_threateningJQJmmLkids It was worthy to explore if abusers’

threatening to hurt kids could predict the amount of violence and the appearance of
appearing lethal violence. (a) Whether threatening to hurt kids could predict the amount
of violence: It was found that the correlation between the two variables was not
significant (F .555, p = .580), poisson regression coefficient was significant (b= -.l32
(exp[-.l32]= .87), p=.000, R-square =.0031), which meant that the previous assumption
could be supported, and the abuser who did threaten to hurt kids had more amount of
violence. (b) Whether or not threatening to hurt kids could predict the appearance of
lethal violence: It was found that the correlation between the two variables was not
significant (Chi-square= .402, p= .526), logistic regression coefficient was not significant
(b= 1.299 , p=.562) (Table 38).

Gallin (1999)

 

Pointed out that most of Taiwan wives in her sample in south Taiwan villages (24 out of

93

25) could not be treated well by their mother-in-law, even about half of them reported to
be bruised by their mother-in—law while living with them. Chen (1997), moreover,
submitted the womb theory, which assumed that the poor relationship between wives and
mother-in-laws might cause the mother-in-laws trigger their sons to abuse their wives.
After collecting the data fiom the 2 centers, the author requested the social workers
the supplementary data on whether the victims lived with their mother-in-laws and had a
bad relationship with them while they were abused. It was found that 31.0% of victims
living in the urban area lived with their mother-in-laws, while 28.6% of victims living in
the rural area did, but the difference did not reach significance (chi-square = .048,
p= .862)(Table 36). This situation was very different from American families, since
normally the American couples did not live with their mother-in—law. It was worthy to
explore if the victims’ living with mother-in-law and having a bad relationship with her
could correlate to or predict the amount of violence and the appearance of appearing
lethal violence of their male partners. (a) Whether the victims’ living with mother-in-law
and having a bad relationship with her could correlate to the amount of violence: It was
found that the victims’ having argument with mother-in-laws received lower amount of
violence (M= 42.72) than the victims’ having no argument with mother-in-laws(M=

51.88), but this difference did not reach significance. After comparing the victims who

94

 

had no argument with mother-in-laws, argued and lived with mother-in-laws, and argued
but not lived with mother-in-laws, it was found that the means of amount of violence
among them did not reach significance (F: 1.234, p = .295). Therefore, the assumption of
womb theory that the victims’ having higher conflict with mother-in-law triggered the
higher amount of violence could not be supported. (b) Whether the victims’ living and
having bad relationship with mother-in-laws could correlate the appearance of lethal
violence: It was found that the relationship between the victims with and without
argument with mother-in—laws and the appearance of lethal violence did not reach
significance (chi-square = .273, p = .303). After comparing the victims who no-argument
with mother-in-laws, argued and lived with mother-in-laws, and argued but not lived with
mother—in-laws, it was found that the means of amount of violence among them did not

reach significance (chi-square = 2.487, p = .288) (see Table 37).

Table 36 Whether the Victims Lived with their Mother-in-laws in Urban and Rural Areas

urban/rural * cohabit with mother in law Crosstabulation

 

 

 

 

 

 

 

 

\ cohabit with mother
in law
no yes Total

urban/rufiTT urban C0unt 60 27 87

% within urban/rural 69.0% 31.0% 100.0%

rural Count 15 6 21

% within urban/rural 71.4% 28.6% 100.0%

Total Count 75 33 108

% within urban/rural 69.4% 30.6% 100.0%

 

 

 

N91;- chi-square = .048, p== .862

95

 

 

Table 37 Whether Victims’ Living and Arguing with Mother-in-laws Correlate the
Amount of Violence and Appearance of Lethal Violence

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Argue w/ mother-in-law Argue w/mother-in—law
Yes no yes No
(n=36) (n=85) (n=85)
Brief CTS 42.72 51.88 Live with Not live with
(n=l4) (n=22)
27.93 52.14 51.88
T/F test t=.812 F=1.139
p= .413 if 324
Appearing No 5 7 12 10 15 46
lethal Yes 44 5 4 7 39
violence
Chi-square Chi-square = .273 Chi-square = 2.487
p= .303 p= .288

 

 

Table 38 Six Other Variables’ Correlation and Prediction on the Amount of Violence
and Appearance of Lethal Violence

 

 

 

 

 

 

 

Predictor \ outcome variable Amount of violence Appearance of lethal violence
1.time % of being abused r=.003, p=.975 t= .917, p= .361
after cohabitation b=.003, p=.690 b= -.221, p= .517
2.Victim’s assessing on r=.267, p=.003" t= 1.718, p= .088
chance of future re-abuse b=.079, p= .092 b= .148 , p= .030* (R-square
= .030)
3.Abusers’ stalking behavior r=-.016, p=.879 Chi-square=2.052 p= .358
b=-.027, p=.106 b= 1.36, p= .208
[b= .198, p=.001**, R-
square=.034, while using
poisson regression to predict
times of lethal violence]
4.Abusers’ threatening to t= .182, p=.856 Chi-square=.057, p= .811
hurt victims’ family of b= -.408, p=.127 b= 1.096, p=.811
0113i}
5.Abusers’ threatening to r= .555, p= .580 Chi-square=.402, p= .526
hurt the child b= -.132 (exp[-.132]=.88), b= 1.299, p= .562
p= .000*** (R—square
=.0031)
6. living and having bad t= .237, p= .303 Chi-square=2.487, p= .288

relationship with mother-
in-law

 

 

b= .649, p=.l34

 

Note. (1) Exp meant exponential. See Long (1997). (2) *<.05, **<.Ol, ***<.001

96

 

 

 

 

After doing the forward stepwise poisson regression by adding the above first 5

items in the 12 valid DA items to predict the amount of violence, it was also found that

the R-square was from .2916 to .4651. However, the time percentage of abuse after

cohabitation was excluded, since it decreased the R—square. Moreover, the slopes of the

times of stalking, threatening to hurt her family of origin, and threatening to him the kids

were all positive and significant, which meant the higher of them the higher amount of

violence. This result would be more credible than the regression with individual

independent variable (see Table 39).

Table 39 Forward Stepwise Poisson Regression Analyses on the 12 DA Items to
Predict the Amount of Violence in Brief CTS

 

 

 

 

 

 

 

 

 

 

 

 

 

R-square Stepwise b-value p Add or
change in Exclude
R-square
12 DA Items (excluding Item .2916
5, 8, &10)
(Forward stepwise by adding
the following 5 items)
l.Percentage of abuse after .2876 -.0040 -.037 .000 Exclude
cohabitation ***
2.Chance of future re-abuse .3100 .0224 .055 .000 Add
assessed bflictims exp[.055]=1.06) ***
3.Time of stalks .3712 .0612 -.155 .000 Add
(exp[-.155]=.86) ** *
4.Threaten to hurt her family .4405 .0693 -.493 .000 Add
of origin (exp[-.493]=.61) ***
5.Threaten to hurt kids .4651 .0246 -.321 .000 Add
***

(exp[-.321]=.73)

 

Note. (1) Exp meant exponential. See Long (1997); (2) *** p<.001

97

 

that

dist

goc

had

11111

1:11

i“- 57’ ‘87‘7

/ rU/fi‘

c ....j . .0

13.1 1‘

As mentioned in the literature review, Holtzworth-Munroe et al (1994) identified

that whether the batterers were violent outside was one of the three dimensions to

distinguish the different types of male batterers. Based on Holtzworth-Munroe et al

(1994), the author tried to learn whether or not the generality of violence could be the

good criterion to distinguish the male batterers. It was found that generality of violence

had a significant positive correlation with six items in the DA. They included forced sex,

threatening to kill, belief of being killed, alcohol abuse, controlling daily life, and violent

jealousy (Table 40).

Table 40 Correlation Between Whether Violent Outside and Other DA 14 Items

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Whether Violent Outside
DA_l DA_2 DA_3 DA_4 DA_S DA_6(1) DA_6(2)
increase increase choking Forced sex drug abuse threat to belief of
frequency severity kill being
of - of killed
violence violence
1- .462 -.008 .042 .286 .075 .412 .297
.648 .934 .651 .003“ .466 .000* * * .001M
Whether Violent Outside
DA_7 DA_8 DA_9 DA_lO DA_ll DA_12 DA_13
alcohol control assault Violent she wanted he wanted Violent
abuse daily life when jealousy to suicide suicide to child
_ pregnant
L .361 .259 .127 .257 .076 .152 .048
2 .000*** .004" .178 .005” .407 .098 .602

 

 

 

 

 

 

 

 

Note. (1) All above variables were regarded as continuous variables for testing their bivariate

correlation, even thought they were dummy variables. (2) *<.05, **<.01, ***<.001

98

Moreover, afier comparing means on the brief CTS, the DA, and times of lethal
violence between the abusers with and without outside violence (DA Item 14), it was
found that the abusers with outside violence had significantly higher amount of violence
then the abusers without outside violence. Such situation also happened on the DA scores,
the mean difference between the two groups was 2.89, which meant that about 2 other
items were responded positively in addition to the item of presence of outside violence.
However, it was found that the two groups could not reach significant difference on the
times of lethal violence, which might be predicted mainly by the five other items in DA
(Item 1, 3, 4, 6(1), and 12) shown in the previous session (Table 41). Therefore, it might
be said that whether abusers’ presenting outside violence could be a good criterion to
distinguish the subtypes of male batterers, but might not be a good predictor for

predicting the lethal violence.

Table 41 Comparing Means on Brief CTS, DA, and Times of Lethal Violence between

the Abusers with and without Outside Violence

 

 

 

 

 

 

 

 

 

 

 

 

Mean T p
Brief CTS Had outside violence 69.60 -2.024 .049*
No outside violence 42.31
DA Had outside violence 8.11 -S.306 .000***
_ No outside violence 5.22
Time of lethal Had outside violence 1.26 -1.035 .303
Elence No outside violence .86

 

99

 

After doing Ward’s method of hierarchical cluster analysis with squared Euclidean
distance among all respondents to the 15 DA items and setting the cluster solution from 2
through 5, it was found that the 4 clusters best fit the data.

After using the stepwise discriminant analysis, it was found that 96.4% of total
original grouped cases were correctly classified in the predicted group membership.
Among them, 97.7% of original grouped cases were correctly classified in Cluster 1,
94.1% in Cluster 2, 94.7% in Cluster 3, and 96.6% in Cluster 4 (Table 42). However, it
was noted that nine respondents were not grouped in this cluster analysis.

Based on the demographic data, it was found that the differences of abusers’ age,
levels of education, and employment status among 4 clusters could not reach significance.
However, it was found that Cluster 1 had the lowest scores in the amount of violence,
times of lethal violence, the DA scores, and outside violence, so it could be named “low-
violent assaulter”. Cluster 2 had medium scores in the amount of violence, and lethal
violence, but had the highest scores in increased frequency of violence, increased severity
of violence, and violence to children. Cluster 2 also had the lowest scores in control daily
life and violent jealousy, so it would be appropriate to name it as “medium-violent, and

low-controlling assaulters”. Cluster 3 had the highest scores in amount of violence, lethal

100

violence, choking, forced sex, alcohol abuse, and assault when pregnant. It was also
found that all batterers in Cluster 3 had alcohol abuse problem, so it would be appropriate
to name them “high-violent, high-lethal, and alcoholic assaulters”. Cluster 4 had the
second highest scores in amount of violence and the times of lethal violence, and also had
the highest scores in the DA total, threat to kill, control daily life, violent jealousy, threat
to commit suicide, and violence outside. It was also found that all of batterers in Cluster 4
had threatened to kill her and controlled her daily life. Cluster 4 was to be named as
“high-violence, high controlling assaulters” (Table 43).

Regarding years of abuse, it was found that the Cluster 3 had the longest history of
abuse after cohabitation (M= 12.53 years), and the differences among 4 clusters reached
significance (F = 2.892, p= .039). However, the difference of the time percentage of
abuse after cohabitation did not reach significance (F=— .917, p=.435), but the it was found
that the Cluster 4 started to abuse their partners before cohabitation (M= 126%), which
may mean that their high level of controlling convinced their female partners that the
abusers did love them, even though they had the high amount of violence. Regarding the
victims’ assessment on chance of being re-abused in the future, the differences among the
4 clusters reached significance and Cluster 1 was the lowest. The four clusters did not

Show difference on the stalking behaviors (chi-square = 7.514, p= .584). It might be

101

because this item categorized the times (such as none, 1-to 2 times, at least 3 times, or
don’t know), and lost the real numbers the victims experienced or it might be because
stalking behaviors did not difl'erentiate the 4 clusters of intimate assaulters. It was found
that threat to hurt the victims’ family of origins to prevent victims’ leaving could
significantly differentiate the 4 clusters (chi-square = 9.962, p= .019), among which
Cluster 3 had the relatively higher chance to do it (66.7%). However, it was found that
the 4 clusters could not be significantly differentiated by abusers’ threat to hurt kids to

prevent victims’ leaving (chi-square = 4.233, p = .237).

Table 42 Classification Results in Stepwise Discriminant Analysis

Classification Results ”1°

 

 

 

 

 

 

 

 

 

 

Predicted Group Membership

Ward Method 1 2 3 4 Total
Original Count 4T 46 1 0 0 47
2 1 16 0 0 17
3 0 0 l8 1 19
4 0 0 1 28 29
% 1 97.9 2.1 .0 .0 100.0
2 5.9 94.1 .0 .0 100 0
3 .0 .0 94.7 5.3 100.0
4 .0 .0 3.4 96.6 100.0
Cross-validateda Count 1 42 3 1 1 47
2 1 14 0 2 l7
3 l 2 15 l 19
4 0 0 1 28 29
% 1 89.4 6.4 2.1 2.1 100.0
2 5.9 82.4 .0 11.8 100.0
3 5.3 10.5 78.9 5.3 100.0
4 .0 .0 3.4 96.6 100.0

 

 

 

 

a. Cross validation is done only for those cases in the analysis. In cross validation.
each case is cla35111ed by the functions derived from all cases other than that case.

b. 96.4% of original grouped cases correctly classified.
c. 88.4% of cross-validated grouped cases correctly classified.

102

Table 43 Difference of Variables Among the 4 Clusters on CTS, Lethal Violence, DA,

15 DA Items, and Other Variables

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

four clusters ANOVA
1 2 3 4 F p
Numbers of respondents 47 17 19 29
(38.8%) (14.0%) (15.7%) (13.9%)
gge of abusers 38.76 42.53 41.11 39.66 .773 .512
levels of education (chi-square) 16.454 .561
employment status(chi-square) 8.162 .227
amount of violence ACTS) 26.28 55.41 78.42 71.28 6.360 .001**
lethal violence or not .32 .53 .63 .41 2.119 .102
time of lethal violence .51 .88 2.11 1.28 3.367 .021*
DA 3.89 6.35 8.26 8.62 37.360 .000“
dal. increased freq of violence .40 .71 .26 .66 4.171 .008M
da2. increased severity of .38 .88 .16 .86 17.061 .000M
violence
da3. Choking .32 .59 .68 .34 3.517 .018*
da4. forced sex .36 .35 .74 .59 3.553 .017*
da5. drug abuse .06 .00 .11 .03 .745 .528
da6(l). he threatened to kill .21 .71 .89 1.00 36.906 .000”
da6(2). she believes to be killed .17 .76 .47 .86 20.351 .000”
da7. alcohol abuse .19 .18 1.00 .38 21.014 .000"
da8. control daily life .40 .18 .37 1.00 19.281 .000“
da9. assault while pregnant .30 .18 .74 .24 6.445 .000”
dal 0. violent jealousL .36 .00 .68 .79 14.982 .000"
dal 1. she threatened/tried to .28 .76 .74 .45 7.101 .000M
commit suicide
da12. he threatened/tried to .21 .24 .26 .31 .307 .820
commit suicide
dal3. violent to child .21 .71 .58 .48 6.190 .001"
dal4 violent outside .04 .12 .58 .59 17.621 .000“
Years of abuse afier cohabit 6.49 8.55 12.53 7.35 2.892 .039*
Time % of abuse after cohabit .54 .56 .91 1.26 .917 .435
Victims’ assessing on future re- 4.89 7.06 6.42 7.55 6.302 .001**
abuse in the scale of 0 to 10
Stalking (3 categories, chi- 7.514 .584
square)
Threat to hurt victims’ family of (27.7%) (29.4%) (66.7%) (48.3%) 9.962 .019*
origins or not to prevent victims’
leave(chi-square)
Threat to hurt kid or not to (17.0%) (29.4%) (27.8%) (48.3%) 4.233 .237

prevent victims’ leave(chi-
square)

N01;- (1) Nine respondents were not grouped in the cluster analysis. (2) *p<.05, **p<0.01

103

For exploring the components of abusers’ pathological behavior directly from the
DA, the author conducted the principle component analysis of factor analysis in the 13
DA items related abusers’ behaviors (excluding DA 6(2) “she believed to be killed” and
DA 11 “she wanted suicide”). Four components were identified in this process (see Table
44). Component 1 included alcohol abuse, force sex, choking, threat to kill, and violence
outside, which almost related to violence. Alcohol abuse might not directly relate to
violence, but alcohol abuse of intimate abusers was for self-medication to inhibit violence.
Therefore, alcohol was violent in nature in the intimate assaulter pOpulation (Andrew
Barclay, personal communication, April 17, 2001). Component 2 included control daily
life, violent jealousy, and he wanted suicide, which almost related to insecure behavior,
since intimate abusers’ feeling insecure toward themselves might project to their partners
and then crash together toward the suicidal thought. Component 3 included the increase
severity of violence and increase frequency of violence, which might have a underlying
meaning of the violence feeding on itself, since once the insecure and abusive men used
violence, it might have to use more and more violence to feed their psychological need.
Component 4 included violence to child and assault when victims were pregnant, which

might come from the abusers’ insecure pattern and is similar to animals’ gene control

104

behavior. Animals would kill the newborn animals, which did not contain their gene.
Based on this assumption, the batterers might not believe the fetus or kids might not

contain their gene (Andrew Barclay, personal communication, April 17, 2001).

Table 44 Factor Analysis of Abusers’ Pathological Behaviors in DA Items related to
Abuser’s Behaviors (excluding Item 6(2) and 11)

Rotated Component Matrix ‘

 

 

 

 

 

Component
1 2 3 4

a7_alcohol abuse .645 .129 -.260 .156
da4_force sex .643 .190 -.105
da3_choke .629 -.358 .157 .261
da6(1)threat to kill .526 .232 .225
da14_violent outside .525 .450 - 134 -.118
da8_control daily life .812 .165
da10_violent jealousy .333 .603
da12_he wanted suicide .484
da2 increase sever or
not— .111 .804 .164
dal increase fre or
not— q .119 .762
da13_violent to child .117 .719
da5_drug abuse .140 -.206 -.617
da9 assault when

regnant .197 .133 -.390 .611

 

 

 

 

 

Extraction Method: Principal Component Analysis.
Rotation Method: Varimax with Kaiser Normalization.

a. Rotation converged in 6 iterations.

WW It was found that of the 121 samples, only 10

(8.3%) victims had civil protection orders (no matter temporary or regular protection

order), 34(28.l%) victims were in the protection order application process (temporary

105

protection order only), and 76(62.8%) victims did not apply for the protection order
(Table 45). In Taiwan, the temporary or regular protection orders are civil matters, and it
had to be applied for based on the victims’ willingness, except for the emergent
protection order, which the prosecutor, police officer, and public social worker can
directly apply for. It had to be noted that there was a limitation on any interpretation of
protection order related questions, since the samples were unavoidably focused on newly
Opened cases. The ten samples having temporary or regular protection orders were old

clients experiencing the re-abuse by their male partners in the time period set in this

 

 

study.
Table 45 Whether the Victims Have Protection Order
whether have protection order
Valid Cumulative
Frequency Percent Percent Percent
PVETTd has P0 10 8.3 8.3 8.3
in P0 apply process 34 28.1 28.3 36.7
not apply P0 yet 76 62.8 63.3 100.0
Total 120 99.2 100.0
Missing System 1 .8
Total 121 100.0

 

 

 

 

 

 

 

IhLmrceptionnLeffchoLptotectionnrders. Of the 10 victims owning the

protection orders, 6 (60%) of them responded that the protection order was effective in

preventing their being re-abused by the abusers, 3 (30%) of them responded it was “so

106

so” in its effectiveness, and 1 (10%) of them responded it was not effective. It was found
that there was no significant difference in the CTS, the DA, and the times of lethal
violence among victims’ different perceptions on effectiveness of protection order (Table
46). However, it was found that the victims who experienced no lethal violence
responded that the protection order was effective, while the victims experienced lethal
violence responded that the protection order was not effective or “so so”, which
confirmed its significant correlation after testing chi-square (chi-square = .6528, p
= .038). It could be related to the learned helpless, which comes from experiencing the
serious domestic violence and from being exposed to domestic violence for a long period
of time which makes the feeling increasingly helpless. Learned helpless is one of
important criteria of battered women syndrome pointed out by Walker (1984). On the
other hand, it could be also possible that the male partners were very violent and their
violence could not be stopped by the protection orders. It had to be noted that there was
limitation on any explanation on the effectiveness of the protection order perceived by the

victims, since the samples were mainly focused on newly Opened cases.

107

Table 46 AN OVA on CTS, DA, and Times of Lethal Violence among Victims’ different

perception on effectiveness of protection order

 

 

 

 

 

 

 

 

 

 

 

CTS DA Times of lethal
violence

Victims’ Not effective Mean 60 9 2
perception on Number 1 1 1
effectiveness So so Mean 88.33 8.67 3.67
of protection Number 6 6 6
order Effective Mean 35.67 7.33 0
Number 3 3 3
AN OVA F 1.526 .399 .899
.282 .686 .449

 

 

 

 

 

WWW Of the ten victims

having protection orders, seven (70%) responded their male partners did violate the

protection orders, while three (30%) responded not. When asked the ways of violating

protection orders through multiple items, the previous seven samples had fourteen

responses. There were 14 ways of violating protection orders. Three (42.9% of responded

victim) were re-beaten, 4 (57.1%) were re-threatened, 5 (71.4%) were harassed by phone,

and 2 (28.6%) had come close to the victims. While comparing the means between the

appearance and non-appearance groups on the CTS, the DA, and times of lethal violence,

no significant difference was found (t = .527, p = .635; t=-1.156, p=.331; L723, p=.522,

respectively), which might result because the number of respondents in the sample, 7,

was too small.

108

 

 

WWW Of the total 121 samples, 84

responded with their worries about the protection order in the multiple response format.
The most common worry was that the protection order might stimulate the abusers’ anger
(68.6%), then unsound protection (34.9%), the application process being too slow
(30.2%), worrying about child’s safety (22.1%), and worrying they can not see the child
(14.0%), and others (9.3%). The category “others” included worrying about presenting in
the court, worrying about hurting the relationship, worrying about the abuser’s killing all
the families and himself, worrying about hurting the male abuser, and worrying about

having no living fee (Table 47).

Table 47 Victims’ Worries about the Protection Order in Multiple Response Scale
Worries on protection order count %of response % of clients

 

 

Unsound protection 30 19.5 34.9
Stimulate anger 59 38.3 68.6
too slow 26 16.9 30.2
worry child safety 19 12.3 22.1
worry can’t see child 12 7.8 14.0
others 8 5.2 9.3

Total responses 154 100.0 179.1

35 missing cases; 86 valid cases

 

WWW When asked whether or not they

had suggestions (up to two) about the protection orders, 15 victims provided 20

109

 

 

 

the 3,

ther

that

prot

proc

app

 

lim-

wit

off

suggestions. Of the twenty suggestions, four were suggestions on decreasing the time for
the judicial review during application process for protection order, two were regarding
the necessity of the treatment program for the batterers, and another two were regarding
that child visitation had to be warranted or the child custody had to belong to the victims.
The rest of the suggestions were suggested once each. They were that the term. of
protection order is not long enough, protection orders should connect to the divorce plea
process, victims should be allowed to waive the right to appear in court during
application process, arresting the abuser if he violated the protection order should not be
limited only when the abuse was witnessed by the police and only when the abuse was
witnessed by the third party, and the protection order should be applied for actively by the

official agencies.

110

CHAPTER 5 DISCUSSION

3 E l E. 1.

The purpose of this study was to see whether or not the Danger Assessment (DA)
scale and some other variables could be used to accurately predict the amount of violence
as well as the lethality of intimate assault cases in Taiwan, and to preliminarily explore
the batterer typology in Taiwan intimate assaulters. The results of testing the four
hypotheses of this study organized as the following.

(1) The correlation between the scores of the DA and the amount of violence in the Brief
CTS reached significant level. By using poisson regression, it was found the DA
could explain 29.26% of total variance in the amount of violence on the Brief CTS.
However, it was found that three items in the DA, Item 5 (drug abuse), Item 8
(control daily life), and Item 10(violent jealousy), could not significantly predict the
amount of violence in the Taiwan sample.

(2) In the predictive accuracy test of predicting lethal violence, the area under the ROC
curve of the DA was .753, which was moderate and satisfactory accuracy, and the

area under the ROC curve of the Brief CTS was .718, which was a lower moderate

lll

 

susr

fou:

3131

(3) 01

and still a satisfactory accuracy. It was found that the correlation coefficient between
lethal violence and the DA was .364 (p=.000), and the correlation coefficient between
lethal violence and the Brief CTS was .333 (p=.000). While using 15 DA items to
predict the appearance of lethal violence in logistic regression, it was found that the
15 DA items could explain 39.83% of total variance in the appearance of lethal
violence. However, a S-item Brief DA was established since the lS-item DA was
suspicious to. have the collinearity problem while predicting lethal violence. It was
found that the 5-Item Brief DA could explain 38.65% of total variance in the
appearance of lethal violence.

(3) Of the 6 other important variables (they were time percentage of being abused after
cohabitation, victims’ assessment on the chance of future re-abuse, abusers’ stalking
behavior, abusers’ threatening to hurt the victims’ family of origin to prevent victims’
leaving, the abusers’ threatening to hurt the kid to prevent victims’ leaving, and
victims’ arguing with mother-in—laws), it was found that only victims’ assessment on
the chance of future re-abuse and the abusers’ threatening to hurt the kid to prevent
victims’ leaving could significantly predict the amount of violence. Moreover, it was
found that only victims’ assessment on the chance of future re-abuse could

significantly predict the abusers’ appearance of lethal violence. It was also found that

112

the victims’ having argument with mother-in—laws did not have a significant
relationship with the amount of violence, so the previous assumption that the victims’
having higher conflict with mother-in-law triggered the higher amount of violence
could not be supported.

(4) In cluster analysis, it was found that the 4 clusters of intimate assaulters could be
identified in Taiwan sample. They could be named as “low-violent assaulters”,
“medium-violent, and low-controlling assaulters”, “high-violent, high-lethal, and
“alcoholic assaulters”, and “high-violence, high controlling assaulters”.

In addition, in factor analysis in the 13 DA items related abusers’ behaviors
(excluding DA 6(2) “she believed to be killed” and DA 11 “she wanted suicide”), four
components were identified. Component 1 included alcohol abuse, force sex, choking,
threat to kill, and violence outside, which almost related to violence. Component 2
included control daily life, violent jealousy, and he wanted suicide, which almost related
to insecure behavior. Component 3 included the increase severity of violence and increase
frequency of violence, which might have an underlying meaning of the violence feeding
on itself. Component 4 included violence to child and assault when victims were pregnant,
which might come from the abusers’ insecure pattern and is similar to animals’ gene

control behavior.

113

- . .... . . O O . I .
0W4'L0 H -..,Htt .1 \ .s .111 t -.. (IUJ- l._ll_o. :- .

As mentioned earlier, even till now there is no concrete theoretical approach for
identifying risk marker of offenders. However, Quinsey et a1 (1998) argued that there is
no reason why theoretically relevant predictors could not be used in an actuarial model.
Based on the findings of this study, the author proposed a theory to help better understand
the risk assessment on the behavior patterns of male intimate abusers in Taiwan. This
proposed theory was named the pathological behavior appearance theory, which meant
that the appearance of different kinds of pathological behaviors could be the traits of
different types/clusters of intimate assaulters and, moreover, some pathological behaviors
could be clustered. Furthermore, some specific portions of abusers’ pathological
behaviors could be good predictors for the amount of intimate violence (such as all DA
items, except drug abuse, control daily life, and violent jealousy) and for the appearance
of lethal violence during the intimate assault (such as increasing the frequency of
violence, choking, he threaten to kill, he threatened/tried to commit suicide) or even for
the future recidivism of intimate assault (such as the predictors used in Spousal Assault

Risk Assessment, see Kropp, Hart, Webster, & Eaves (1999)).

114

Moreover, as found in this study, the assumption of womb theory that the victims’
having higher conflict with mother—in-law triggered the higher amount of violence could
not be supported. Maybe the womb theory could be a good theory to explain the
phenomenon of how intimate assault happened in some Taiwan families, but not all
Taiwan families, nor for the risk assessment on amount of violence and lethal violence in
Taiwan intimate assault cases. There might be some other important variables existed and
good for risk assessment of intimate assaulters. Based on this study, the different clusters
of male intimate assaulters in pathological behavior patterns were preliminarily identified.
It was found that whether the outside violence appeared could be the criteria to
differentiate the clusters among Taiwan male intimate abusers found in this study.
Moreover, it was found that the amount of violence in the Brief CTS, and 13 DA items
(excluding drug abuse and abuser’ threat to suicide) were significantly different among
the 4 clusters. However, the further typology study of male intimate abusers based on

pathological behaviors among male intimate abusers would still be encouraged.

‘qlll‘l it...” 0 -Ii,“t 0’ o 1114';
Healey, Smith, and O'Sullivan (1998) conducted a nationwide study for evaluating

the intimate abuser treatment programs in the US. They concluded that “one-size-fits-all”

115

intervention can not accommodate the diverse population of batterers, so new directions
of treatment programs had to be encouraged. One of the new directions was that the
intervention has to be based on the batterer typologies, such as longer supervision and/or
treatments might be required for the higher risk offenders. Even though that, does that
mean the prolonged treatment can eventually cure the higher risk intimate assaulters?
Based on his clinical experience, Dutton (1995a, p174) pointed out that for batterers with
borderline personalities, more thorough or longer treatment is probably required, and for
batterers with psychopathic personalities, such as vagal sign reactors (those who become
internally calmer and more focused while they batter their wives), treatment may be
unproductive.

Dutton (1995b, p250), moreover, argued that in order to devise a form of therapy
to be used as a condition of probation for men convicted of intimate assault, certain sets
of requirements must be met, such as the therapy had to be compatible with criminal
justice philosophy, which emphasizes personal responsibility for action. Then, he argued
that traditional feminist treatment programs, such as the Duluth model (noted by author),
might not work well on the high risk intimate abusers, because they might not change
easily by gender equality education or be easily scared by criminal justice intervention.

He proposed that cognitive-behavioral therapy, which is based on the social learning

116

theory, might be a good way to intervene and help the intimate abusers stop their violence,
because the social learning theory not only acknowledges the formative role of situational
events in shaping habit patterns, but also stresses choice and responsibility for individual
actions. Moreover, Paul Gendreau (1996), a Canadian correctional psychologist,
submitted that relapse prevention is one of the eight principles of effective intervention
with offenders. Relapse prevention (RP) was developed by Marlatt and Gordon (1985)
and was originally for helping the substance abuse clients to keep sober. The rationale of
RP was using cognitive-behavioral therapy to help the client identify relapse/reoffense as
a cycle underlying the clients’ behavioral pattern and eventually to disconnect it by
making better choices about every thought and behavior.

However, if the criminal justice system treats people convicted of similar crimes
differently it might raise questions about justice. Therefore, three crucial mechanisms
would need to be developed, and they are expert witness, risk assessment scale, and
prolonging treatment in an appropriate manner. The expert witness may be the
psychologists, psychiatrists, counselors, social workers, or any other mental health
professionals, who can address the batterers’ specific problems and assess their risk based
on their clinical experience. However, it would be much more accurate if the clinicians

can use actuarial risk assessment scales to assess the potential risk posed by the offenders

117

(Hanson, 1997). Regarding the prolonging treatment for the higher risk batterers,
treatment programs in Colorado’s 18'“ Judicial District, close to Denver, provide a good
example in which the treatment providers can prolong the treatment term based on the
support of probation officers and judges if they found the batterer was a risk in their
checklist (Healey, et al 1998). However, one could challenge that putting the prolonged
batterers in the same group as the lower risk batterer might possibly cause the prolonged
batterers to be angry and unproductive if they saw the low risk ones were allowed to
leave earlier than them. The Total Health Education Counseling Center, Lansing,
Michigan, provided another good example of dealing with this problem as the author
observed. After participating in the 26-week therapeutic group with relapse prevention
and maturity model, the batterers could be transferred to the graduate groups, which
required them to participate 1 or 2 times a month based on the probation order till their
probation terms finish. The members in graduate groups can present the recent situations
they have been in, and then discuss with the counselor or the other members alternative
ways to deal with the situations based on the relapse prevention language. During the
time of prolonged treatment, the treatment providers could observe whether the clients
make progress and, as the author observed, most of the batterers showed much more

willingness and felt much more comfortable while in this group. In the other words,

118

based on the attachment theory, they can try to build a secure attachment with the
treatment providers, which eventually leads to progress, even though some of them might

naturally have an insecure pattern of attachment.

AdxantageacfthiaSn-idy

There were some advantages of this study as the following. First, this study could
introduce the American risk assessment scale and variables to Taiwan, and help to
understand the feasibility of using them in Taiwan. Second, multiple predictive accuracy
tests were used in this study to help gain more knowledge about the predictive accuracy
of the DA and the Brief CTS. Finally, by using cluster analysis and factor analysis, it was
found that some important components and clusters were discovered, which did help the

researcher to learn more about the nature of different types of male intimate assaulters.

111° 1 E 1 . S 1
There were some disadvantages of this study as the following. First, this study did
not involve the male partner’s side of the story. It is reasonable to assume that the female

victim’s sides of the stories are subjective and biased. However, for ethical reasons and

the reason of victims’ safety, it may be better not to pair the abusers in this study. Second,

119

verbal aggression did not be included in the measurement of intimate abuse. It is
reasonable to assume that verbal aggressions and physical aggressions from abusive
partners are equally harmful to the abused intimates. However, since the Danger
Assessment (DA) is focused on physical aggression, this study would only focus on
physical aggression. Third, it might be worthwhile to explore whether the male abusers
had a mental problem, such as s psychotic disorder, paranoid disorder, or even personality
disorders, which were all included in the Spousal Assault Risk Assessment, SARA, a risk
assessment that mainly targeted the recidivism risk factors (Kropp, Hart, Webster, Eaves.
1999). However, there were problems with collecting this information while they
collected from the victims because the victims might exaggerate the abusers’ mental
problem (Holtzworth-Monroe, personal communication, August 6, 2000) and the local
social workers might have limited training to recognize the abusers’ mental problems and

also they might have a limited chance to contact the abusers.

SuggestinnachEunirLSnidx
For the future study, the author sincerely considered the following suggestions.

First, since there were two different types of risk assessment identified in the criminal

justice field. One was risk assessment for lethality and the other was risk assessment for

120

the recidivism. Prediction for the abusers’ recidivism would be very important in judicial
decision making, in addition to lethal risk assessment. It was suggested that the recidivist
risk assessment should be comparably emphasized in the research arena and practical
arena. Second, this study did not explore more of the psychological characteristics of
male intimate abusers, because the samples were all focused on female victims. The
better research sample for understanding the psychological characteristics of abusers
would be at the batterer treatment settings. It was suggested that it would be better to
conduct the research during the beginning phase of the treatment or by the time of
starting treatment. Otherwise, during and after the treatment, the abusers might have
some changes in their original thinking patterns or psychological patterns. Third, future
research was encouraged to learn more about the typology of male intimate abusers in
Taiwan based on the cluster analysis. Since it would be easier for social service agents
and police officers to identify the different types of inmate abusers and help to make a
safe and swift decision while dealing with the intimate assault cases. Fourth, in Taiwan,
the temporary or regular protection orders are civil matters, and it has to be applied based
on the victims’ willingness, except the emergent protection order, for which the
prosecutor, police officer, and public social worker can directly apply only in very

emergent situations. Recently, it was discussed by Taiwan feminist groups that it was

121

better to change the civil protection order into the criminal protection order. Otherwise, it
was a long process, about 2-3 months, to turn a civil protection order into criminal
procedure, when the abusers violate the civil protection order. Based on the findings of
this study, it was found that 68.6% of clients worried about the stimulation of abusers’
anger if they applied for the protection order. If the protection order could be turned to
criminal matters, the prosecutors and the probation/parole officers could be directly
involved in the intimate assault cases, which could be a good solution to provide more

sound protection as well as a swift avenue to prevent the abusers’ re-abuse.

122

APPENDICES

Appendix 1

Conflict Tactics Scale-Form R (CTS— Form R)

Subject ID#
[Directions] It is normal for couple or intimate partners to have conflict. There are a lot of
different things that you‘and your partner can do when you both have conflict or
disagreement. We would like you to try and remember what your [male partnerj went on
during the past year when both of you had disagreement. Please circle a number for each
of the things listed below to show [how often your male partner did it IN THE PASl]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1Presenttime(s) in the past year
0=never
l=once
2.=twice
3=3-5times
4=6-10times
5=1 l-20times
6=more than 20
1. discussed an issue calmly 0 1 2 3 4 5 6
2. got information to back up his side of things 0 1 2 3 4 5 6
3. brought in, or tried to bring in, someone to help settle 0 l 2 3 4 6
things
4. insulted or swore at you 0 l 2 3 4 5 6
5. sulked or refused to talk about an issue 0 l 2 3 4 5 6
6. stomped out of the room or house or yard 0 1 2 3 4 5 6
7. cried 0 1 2 3 4 5 6
8. did or said something to spite you 0 1 2 3 4 5 6
9. threatened to hit or throw something at you 0 1 2 3 4 5 6
10. threw or smashed or hit or kicked something 0 l 2 3 4 5 6
ll. threw something at you 0 l 2 3 4 5 6
12. pushed, grabbed, or shoved you 0 l 2 3 4 5 6
13. slapped you 0 1 2 3 4 5 6
14. kicked, bit, or higou with a fist 0 l 2 3 4 5 6
15. hit or tried to hit flu with something 0 l 2 3 4 5 6
16. beat you_up (beat me continuously in one episode) 0 l 2 3 4 S 6
17. choked you 0 l 2 3 4 5 6
18. threatened you with a knife 0 1 2 3 4 5 6
19. used a knife or fired a gun 0 1 2 3 4 5 6

 

 

 

 

 

 

 

 

 

blot:- from Straus, M. (1990a). Measuring intrafamily conflict and violence: The Conflict
Tactics (CT) Scales. In M. A. Straus & R. J. Gelles (Eds.) Physical violence in American
families: Risk factors and adaptations to violence in 8, 1 45 families (pp.29-4 7). New
Brunswick, NJ: Transaction Publishers.

124

 

Appendix 2

Danger Assessment (DA)

Subject ID#

 

Please answer the following questions “yes” or “no”

__ 1. physical violence increased in frequency

__ 2. physical violence increased in severity

__ 3. partner tried to choke victim

__ 4. partner forced victim to have sex

__ 5. partner used street drugs

__ 6(1). Partner threatened to kill victim2 ,

__ 6(2) Victim believes partner is capable of killing her

__ 7. partner is drunk every day

__ 8. partner controls all victim’s activities

__ 9.partner beat victim while pregnant

__ 10.paitner is violently jealous of victim (says things” if I can’t have you, no one
can.”

__ 11.victim threatened/tried to commit suicide

__ 12.partner threatened/tried to commit suicide

__ 13.partner is violent toward victim’s children

__ l4.partner is violent outside house

 

2 The original Item 6 is “Partner threatened to kill victim or victim believes partner is capable of killing
her”. Originally, there are 15 items in DA, but Item 4 was deleted, since Taiwan is a firearm-control country.
Original Item 4 is “a gun is presence in the house”.

125

 

Appendix 3

Brief Conflict Tactics Scale (Brief CTS)

Subject ID#

[Directions] It is normal for couple or intimate partners to have conflict. There are a lot of
different things that you. and your partner can do when you both have conflict or
disagreement. We would like you to try and remember what your bale partneij went on
during the past year when both of you had disagreement. Please circle a number for each
of the things listed below to show 110w often your male_partner did it IN THE PASTJ

 

 

 

 

Eresent time(s) in the past year

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0=never
l=once
2.=twice
3=3-5times
4=6-10times
‘ 5=11-20times
6=more than 20
I. threw something at me 0 l 2 3 4 5 6
2. pushed, grabbed, or shoved me 0 1 2 3 4 5 6
3. slapped me 0 1 2 3 4 5 6
4. kicked, bit, or hit me with a fist 0 l 2 3 4 5 6
5. hit or tried to hit me with something 0 l 2 3 4 5 6
6. beat me up (beat me continuously in one episodeL 0 1 2 3 4 5 6
7. choked me 0 1 2 3 4 5 6
8. threatened me with a knife 0 l 2 3 4 5 6
9. used a knife or any killing behavior except choking 0 l 2 3 4 S 6

 

 

 

126

 

 

Appendix 4

Victim Questionnaire

 

 

 

Number of Subject Interviewer’s Name Agency
Date of filling
1. Demographic Information:
(1) Year of Birth (age )
(2) Place of Birth (city/county)

 

(3) Current habitation

(4) Culture or Ethnicity (according to your father): _Min-nan; _Hakka;__First
generation from Mainland China; _Second generation from Mainland China;_
Third generation from Mainland China;_ Aboriginal (specify ); others
(Specify__)

(5) Educational level:_ (a) illiterate; (b)elementary school level; (c) junior high
school; (d) senior high school; (e) community college; (f) four-year university; (g)
graduate school

(6) Religion:

_CD I do believe in religion (on ); ®I do not believe in religion

 

 

 

2. Information about ABUSER

(1) Is he your (Dhusband; ®cohabiting partner; @other

(2) How old is he?

(3) Place of birth?

(4) Current habitation

(5) Culture or Ethnicity (according to your father): _Min-nan; _Hakka;____First
generation from Mainland China; _Second generation from Mainland China;_
Third generation from Mainland China;_ Aboriginal (specify ); others
(Specify__)

(6) Educational level:___ (a) illiterate; (b)elementary school level; (c) junior high
school; ((1) senior high school; (6) community college; (1) four-year university; (g)
graduate school

(7) Religion:

_____CD He does believe in religion (on ); @He doesn’t believe in religion

 

 

 

 

(8) Current employment status: 0) unemployment (for how long
year____month);®part time employee;® part time employee

 

3. History of Intimate relationship
( 1) Your current marital status: (Dmarried;®divorce;®married and living
separately;@single and live with boyfiiend; ©other
(2) How long have you live with your this current partner __ year( month)

 

127

 

 

4. Employment and Economic situation:
(1) what is your current employment status: @housewife; @unernployment (for how
long: ___year __month;® part time employee; @ part time employee
(2) What is the month family income in your family with your partner? ®below
20000; ©20000-40000; @40000-80000;®80000—1 00000; ®over 100000
5. How likely would you say that your partner would become violent with you during the
dispute in the next year? (Please circle one number which is close to your idea.)

(0: no chance I (10:it’s sure
to happen) to happen)
0 l 2 4 5 6 7 8 9 10

3
L l l l l l l l l l 1

6. Did your partner stalk you after you both having a bad relationship?—
®3 or more than 3 times during last year;®one or two times during last year;®never
@ not that I know of.
7. Relationship with Your Mother-in-law
(1) Do you live with your mother-in—law? __ Yes ; _ No
(2) How does your mother-in-law like you? _ like you very much; __ like you a bit;
__ so so; _ dislike you; _ dislike you very much
(3) To what degree do you think your relationship with your mother-in-law trigger
your being abused by your partner? __ very much; _much; __so so; _not very
much; __not at all
8. Civil Protection Order
(1) Do you have a civil protection order?
_ Yes (if yes, further answer ..) Do you think it is efiective to protect you? ___(Dvery
effective®somewhat effective ® so so G) somewhat ineffective ®not at all
___In the applying process
_ No
(2) What is/would be your concern in applying the civil protection order? (pick one or
more)
1:] the protection is not sound enough
1:1 applying protection order may stimulate your partner’s anger
[:1 it takes to much time to apply
[:1 the protection is not good enough
Cl worry about kids’ safety
Cl worry on not being able to see the kids
El others
(3) Do you have any suggestion for the civil protection order? (Please write down if you
have any suggestion.)

 

 

®

 

128

 

Appendix 5 Consent Form for Victim

Dear Perspective Participant:

This research is to reassess the severity risk markers of intimate abuse among two American Tools in
Taiwan intimate abuse sample. The tools are the Danger Assessment (DA) and the Brief Conflict Tactics
Scale (CTS). In addition to both of the above, there is a Victim Questionnaire for you regarding you and
your partner’s information to let us identify some possibly potential risk markers.

The DA may take you about 10 minutes to fill out, the Brief CTS 5 minutes, and the Victim
Questionnaire 15 minutes. The following are the conditions of confidentiality policy of this research. Please
read through the following.

1. Your name would not be linked to your response. No one except the responded social worker is able to
access the identification of your response.
2. If you have any concern regarding this research, you can contact either of the following people.
(1) The names and telephone numbers of principle investigators are
0 Dr. Charles Corley Telephone: 012-1-517-353-5225
Address School of Criminal Justice
Michigan State University
East Lansing, Michigan 48824 U.S.A.
o Min-chieh Lin Telephone: 012-1-517-355-2770
Address School of Criminal Justice
Michigan State University
East Lansing, Michigan 48824 U.S.A.

(2) The name and telephone number of Chair for University Committee on Research Involving Human
Subjects (UCRIHS) at Michigan State University are:
0 Dr. David Wright Telephone: 012-517-355-2180
Address University Committee on Research Involving Human Subjects
246 Administration Building
Michigan State University
East Lansing, Michigan 48824 U.S.A.
3. You would indicate your voluntary agreement to participate by completing and
returning these three tools. You may decide not to answer any question or not to
respond to all instruments. You can keep this form and contact either one of the above

persons, if you have any concern.

129

 

Appendix 6 Consent Form for Social Workers
Dear Perspective Responded Social Worker: I
This research is to reassess the severity risk markers of intimate abuse in Taiwan intimate abuse
sample by the risk markers supported by studies in north American. We may need your help to assist the

intimate assault victims to fill out either in person or by phone. It may take you or the victim about 30

minutes to fill them out. Be aware some psychological uncomfortable reaction may reveal among the your

client. Please give empathy to them or pacify them if necessary or stop to continue if necessary.
The following are the conditions of confidentiality policy of this research.

1. The identification of victims’ response is confidential. No one except the responded social worker is
able to access the identification of victims’ response.

2. If you have any concern regarding this research, you can contact either of the following people.

(1) The names and telephone numbers of principle investigators are
0 Dr. Charles Corley Telephone: 012-1-517-353-5225
Address School of Criminal Justice
Michigan State University
East Lansing, Michigan 48824 U.S.A.
O Min-chieh Lin Telephone: 012-1-517-355-2770
Address School of Criminal Justice
Michigan State University
East Lansing, Michigan 48824 U.S.A.
(2) The name and telephone number of Chair for University Committee on Research Involving Human
Subjects (UCRIHS) at Michigan State University are:
0 Dr. David Wright Telephone: 012-517-355-2180
Address University Committee on Research Involving Human Subjects
246 Administration Building
Michigan State University
East Lansing, Michigan 48824 U.S.A.

3. You would indicate your volunteer agreement to participate by assisting victims to complete and return
these three tools. You may decide not to help your client to answer any question or not to help your
client to respond to all instruments. You can keep this form and contact either one of the above persons,
if you have any concern.

. . . C
.1.- . I. . .‘..c.0_..u.. . '. . '3 " q“ . 0. ..'.. ____.___

 

I know I understand the confidentiality policy of Min-chieh Lin’s research from Michigan State
University, and I also guarantee that 1 would not release any piece of the response that victim made in the
three tools to anyone.

Signature of Social Worker Date

 

 

130

BIBLIOGRAPHY

131

 

BIBLIOGRAPHY

Aldarondo, E., & Sugarman, D. B. (1996). Risk marker analysis of the cessation and
persistence of wife assault. Journal of Consulting and Clinical Psychology, 64(5),
1010-1019.

American Psychiatric Association. (1994). Diagnostic and statistical manual of mental
disorders (4‘h ed.) Washington, DC: Author.

Berk, R. A., Berk, S. F., & Loseke, D. R. (1983). Mutual combat and other family
violence myths. In D. F inkelhor, R. J. Gelles, G. T. Hotaling (Eds.), The dark side of
families (pp. 197-212). Beverly Hills, CA: Sage.

Bonta, J. (1996). Risk-needs assessment and treatment. In A. T. Hartland (Ed.) Choosing
correctional options that work: Defining the demand and evaluating the supply.
mp.18-32). Thousand Oak, CA: Sage.

Browne, A. (1984). Assault and homicide at home: When battered women kill. In M. J.
Saks & L. Saxe (Eds.) Advances in Applied Social Psychology, 64(5), 1010-1019.

Burgess, E. W. (1928). Can scientific methods be applied to parole administration? In A.
A. Bruce, E. W. Burgess, & A. J. Hamo (Eds.), The working of indeterminate-
sentence law and the parole system in Illinois (pp. 246-249). Chicago: State of
Illinois.

Buzawa, E. S., & Buzawa, C. G. (1990). Domestic violence: The criminal justice
response. Newbury Park, CA: Sage.

Cadsky, O., & Crawford, M. (1988). Establishing batterer typologies in a clinical sample
of men who assault their female partners. Special Issue: Wife battering : A Canadian
perspective. Canadian Journal of Community Mental Health, 7, 119-127.

Campbell, J. C. (2000). Issues in risk assessment in the field of intimate partner violence.

132

Paper presented at the meeting of 5’h Annual BISC-MI (Batterer Intervention Services
Coalition of Michigan) conference, Midland, Michigan. Or [online] available.

  

Campbell, J. C. (1999). Sanction and sanctuary: When battering within cultural contexts.
In D. A. Counts, J. K. Brown, & J. C. Campbell (Eds). To have and to hit: Cultural
perspective on wife beating (op261-285). Urbana, IL: University of Illinois Press.

Campbell, J. C. (1995). Prediction of homicide of and by battered women. In J. C.
Campbell (Ed.), Assessing dangerousness: Violence by sexual oflenders, batterers,
and child abusers (pp96-113). Thousand Oak, CA: Sage.

Campbell, J. C. (1986). Nursing assessment for risk of homicide with battered women,
Advances in Nursing Science, 8(4), 36-51.

Campbell, J. C. (1985). Beating of wives: A cross-cultural perspective. Victimology, 10,
174-185.

Campbell, J. C. (1981). Misogyny and homicide of women. Advances in Nursing Science,
3(2), 67-85.

Carlson, B. (1984). Causes and maintainance of domestic violence: An ecological
analysis. Social Service Review, 58, 569-586.

Chen. J. R. (1997). Spousal abuse in Taiwan. Lectured at the meeting of multicultural
perspective of domestic violence. Michigan State University, East Lansing, Michigan.

Chen. J. R. (1988). The research of marital violence in Taiwan and its therapeutic
strategies. Taipei, Taiwan: Women Research Office, National Taiwan University.

Chen, S. (1997). Regression Analysis. [In Chinese]. Taipei, Taiwan: Whah-Tai.

Chen, W. (2000). History of Taiwan. [Online] available
httpzllwmleksucomlleksufihtrn February 22, 2000-

Chin, Y. (1990). The behavioral analysis of the Chinese and their face and shame [in

133

Chinese]. In K. Young (Ed.) The psychology in the Chinese (pp319-345). Taipei,
Taiwan: Kuei-kuai.

Chou, Y. (1995). Marital violence: Theoretical analysis and social work intervention [in
Chinese]. Taipei, Taiwan: Chu-liu.

Chou, Y. (1993). The research of social support for abused women in Taiwan. [in
Chinese]. Women and Both Genders, 5, 69-108.

Chu, A. (1998). Taiwanese Hakka: History of Hakka in Taiwan. [Online] Available:

1!. AAAA .l,. v. 9" ‘A..‘._s\‘.., 9‘. 1. 3'"

Department of Social Affair, Taiwan Province Government (1993). Women life situation
survey at Taiwan Province. Nantou, Taiwan: Author.

Dobash, R. E. & Dobash, R. (1979). Violence against wives. New York: Free Press.
Dutton, D. G. (1995a). The batterer: A psychological profile. New York: Basic Books.

Dutton, D. G. (1995b). The domestic assault of women: Psychological and criminal
justice perspectives. Vancouver, Canada: The University of British Columbia Press.

Dutton, D. G., & Hart, S. D. (1992). Risk markers for family violence in a federally
incarcerated population. International Journal of Law and Psychiatry, 15, 101-112.

Dutton, D. G. (1988). Domestic assault of women: Psychological and criminal justice
perspective. Boston, MA: Allyn and Bacon.

Fagan, J. A., Stewart, D. K. & Hansen K. V. ( 1983). Violent men or violent husband?
Background factors and situational correlates. In D. F inkelhor, R. J. Gelles, G. T.
Hotaling (Eds.), The dark side of families (pp. 49-68). Beverly Hills, CA: Sage.

Gallin, R. S. (1999). Wife abuse in the context of development and change: A case from
Taiwan. In D. A. Counts, J. K. Brown, & J. C. Campbell (Eds). To have and to hit:
Cultural perspectives on wife beating. (2"‘1 Ed) (pp 252-260). Urbana, IL: The
University of Illinois Press.

134

 

 

Gendreau, P. (1996). The principles of effective intervention with offenders. In A. T.
Hartland (Ed.) Choosing correctional options that work: Defining the demand and
evaluating the supply. mp1] 7-130). Thousand Oak, CA: Sage.

Gondolf, E. W. (1996). Characteristics of batterer in a multi-site evaluation of batter
intevention systems: A preliminary report. Online available:
mammmcamnmnedutpaperslgondolflhatchanhnn or
I 1, . l , ., l 1' .

Gondolf, E. W. (1988). Who are those guy? Toward a behavioral typology of batterers.
Violence and Victims, 3, 187-203.

Goodman, L. A., Dutton, M. A., & Bennett, L. (2000). Predicting repeat abuse among
arrested battered: Use of the Danger Assessment Scale in the criminal justice system.
Journal of Interpersonal Violence, 15(1), 63-74.

Hampton, R. L., Jenkins, P., & Vandergriff-Avery (1999). Physical and sexual violence in
marriage. In R. L. Hampton (Ed). Family violence: Prevention and treatment (2“ ed)
mp168-197). Thousand Oaks, CA: Sage.

Hanson, K.‘& Thornton, D. (1999). Static 99: Improving actuarial risk assessments for
sex offenders. Ottawa, Canada: Department of the Solicitor General Canada. [Online]
available

WWM

Hanson, K. (1997). The development of a brief actuarial risk scale for sexual offense
recidivism (Rapid Risk Assessment for Sex Offense Recidivism, RRASOR). Ottawa,
Canada: Department of the Solicitor General Canada [Online] available

Wm

Hart, B. (1992). Programme standards for batterer intervention services. Reading, PA:
Pennsylvania Coalition Against Domestic Violence.

Healey, K., Smith, C., & O'Sullivan, C. (1998). Batterer intervention: Program
approaches and criminal justice strategies. Washington D. C.: National Institute of

135

 

Justice. Online available:
mamncjrsmgttxtfilesflfififilfim OR
WWW (Adobe file. Open Adobe software first)

Hilberman, E., & Munson, K. ( 1977). Sixty battered women. Victimology, 2, 260-270.

Holtzworth—Munroe, A., Meehan, J. C., Herron, K., Rehman, U., Stuart, G. L. (in press).
Testing the Holtzworth-Munroe and Stuart Batter Typology. Journal of Consulting
and Clinical Psychology.

Holtzworth-Munroe, A., & Stuart, G. (1994). Typologies of male batterers: Three
subtypes and the differences among them. Psychological Bulletin, 116(3), 476-497.

Hotaling, G. T., & Sugarman, D. B. (1986). An analysis of risk marks in husband to wife
violence: The current state of knowledge. Violence and Victims, 1(2), 101-124.

Huang, C., & Chen, J. R. (1997). The causes of sexual oflense and their related issues [In
Chinese]. Taipei, Taiwan: Ministry of Justice.

Kropp, P. R., Hart, S. D., Webster, C. D., Eaves, D. (1999). Spousal Assault Risk
Assessment guide: User's manual. .North Tonawanda, New York: Multi-Health

Systems Inc.
Lee, Y. (1978). Beliefs and culture [in Chinese] Taipei, Taiwan: Chu-liu.

Long, J. S. (1997). Regression models for categorical and limited dependent variables.
Thousand Oaks, CA: Sage.

Maddala, G. S. (1983). Limited dependent and qualitative variables in econometrics.
New York: Cambridge University Press.

Marlatt, G., & Gordon, J. (1985). Relapse prevention: Maintenance strategies in the
treatment of addictive behaviors. New York: Guilford.

McFarlane, J ., Cambell, J ., Wilt, S., Sachs, C., Ulrich, Y., & Xu, X. (1999). Stalking and
intimate partner femicide. Homicide Studies, 3(4), 300-316.

136

McFarlane, J ., Parker, B., Soeken, K. (1995). Abuse during pregnancy: Frequency,
severity, perpetrator, and risk factors of homicide. Public Health Nursing, 12(5), 284-
289.

Mossman, D. (1994). Assessing predictions of violence: Being accurate about
accuracy. Journal ofConsulting and Clinical Psychology, 62, 783-792.

Monahan, J. & Steadman, H. J. (1994). Toward a rejuvenation of risk assessment research.
In J. Monahan & H. J. Steadman (Eds) Violence and mental disorder: Developments
in risk assessment. Chicago, IL: University of Chicago Press.

Monahan, J. (1981). Predicting violent behavior: An assessment of clinical techniques.
Beverly Hills, CA: Sage.

Quinsey, V. L., Hanis, G. T., Rice, M. E., & Cormier, C. A. (1998). Violent oflenders:
Appraising and managing risk. Washington, DC: American Psychological

Association.

Rice, M. E., & Harris, G. T. (1995). Violent recidivism: Assessing predictive validity.
Journal of Consulting and Clinical Psychology, 63, 737-748.

Rosenberg, M. L., & Mercy, J. A. (1985). Homicide and assaultive violence. In US.
Public Health Service (Ed.), Violence as a public health problem (pp. H1—H47).
Atlanta, GA: Author.

Social Affair Department, Taiwan Province Government (1990, 1992). Women life
situation survey in Taiwan Province. Nan-ton, Taiwan: Author.

Sonkin, D. J. (1987). The assessment of court-mandated male batterers. In D. J. Sonkin
(Ed.), Domestic violence on trial: Psychological and legal dimensions of family
violence (pp,l74-l96). New York: Springer.

SPSS (1999a). SPSS base 10.0 application Guide. Chicago, IL: SPSS Inc.

SPSS (1999b). SPSS graduate pack 10.0 for windows [software in CD]. Chicago, IL:

137

SPSS Inc.

Stark, E. & Flitcrafl, A. H. (1985). Spouse assault. In US. Public Health Service (Ed.),
Violence as a public health problem (op. SAI-SA 42). Atlanta, GA: Author.

Straus, M. A., Harnby, Sherry L, S. Boney-McCoy, and D. B. Sugarman. (1996). The
revised Conflict Tactics Scales (CTSZ): Development & Preliminary Psychometric
Data. Journal of Family Issues 1 7(3).283-316

Straus, M. (1991, November). Severity and chronicity of domestic assault: Measurement
implication for criminal justice intervention. Paper presented at the Annual American
Society of Criminology Conference, San Francisco.

Straus, M. (1990a). Measuring intrafamily conflict and violence: The Conflict Tactics
(CT) Scales. In M. A. Straus & R. J. Gelles (Eds.) Physical violence in American
families: Risk factors and adaptations to violence in 8, 1 45 families @p.29-4 7). New
Brunswick, NJ: Transaction Publishers.

Straus, M. (1990b). The Conflict Tactic Scales and its critics. In M. A. Straus & R. J.
Gelles (Eds.) Physical violence in American families: Risk factors and adaptations to
violence in 8, 145 families @p.49- 73). New Brunswick, NJ: Transaction Publishers.

Straus, M. (1990c). Injuries and frequency of assault and “representative sample fallacy”
in measuring wife beating and child abuse. In M. A. Straus & R. J. Gelles (Eds.)
Physical violence in American families: Risk factors and adaptations to violence in
8, 1 45 families 09p. 75-91). New Brunswick, NJ: Transaction Publishers.

Straus, M. (1990d). Ordinary violence, child abuse, and wife beating: What do we have in
common. In M. A. Straus & R. J. Gelles (Eds.) Physical violence in American
families: Risk factors and adaptations to violence in 8, 1 45 families (pp. 403 -423).
New Brunswick, NJ: Transaction Publishers.

Straus, M. (1990e). Appendix B. New scoring methods for violence and new norms for
the Conflict Tactics Scales. In M. A. Straus & R. J. Gelles (Eds.) Physical violence in
American families: Risk factors and adaptations to violence in 8, 145 families
mp. 535 -55 2). New Brunswick, NJ: Transaction Publishers.

138

Straus, M., & Smith, C. (1990). Violence in Hispanic families in the United States:
Incidence rates and structural interpretations. In M. A. Straus & R. J. Gelles (Eds.)
Physical violence in American families: Risk factors and adaptations to violence in
8, 1 45 families (pp.341-367). New Brunswick, NJ: Transaction Publishers.

Straus, M., & Gelles, R. J. (1986). Societal change and change in family violence from
1975 to 1985 as revealed by two national surveys. Journal of Marriage and the
Family, 48, 477-485.

Stuart, B. P., & Campbell, J. C. (1989). Assessment of patterns of dangerousness with
battered women. Issues in Mental Health Nursing, 10, 245-260.

Snyder D. K. & F ruchtman, L. A. (1981). Diflerenfial patterns of wife abuse: A data-
based typology. Journal of Consulting and Clinical Psychology, 49, 878-885.

Taiwan Ministry of Interior (2001). Taiwan population statistics. [Online] Available

_httnzL/mringLM , April 5, 2001-

Tang, C. S. (1999). Psychological abuse of Chinese wives. Journal of Family Violence,
13(3), 299-314.

Tang, C. S. (1994). Prevalence of spouse aggression in Hong Kong. Journal of Family
Violence, 9(4), 347-356.

Tang, C. & Tsai, C. (1997). I am suffering: Going through the shadow of marital violence.
[in Chinese]. Taipei, Taiwan: Teacher Chang.

Tang, H. (1993). The preliminary stuay of abused women in marital violence and their
coping process. Unpublished Masters Thesis. Shuchow University, Taipei, Taiwan.

Tolman, R. M, & Bennett, L. W. (1990). A review of research on men who batter. Journal
of Interpersonal Violence, 5, 87-118.

Walker, L. E. (1984). The battered women syndrome. New York: Springer.

 

Walla

Ml.

Wei

Walker, L. E. (1979). The battered women. New York: Harper & Row.

Wallace, H. (1999). Family violence: Legal, medical, and social perspectives. Boston,
MA: Allyn and Bacon.

Weisz, Tolman, and Saunders (2000). Assessing the risk of severe domestic violence: The
importance of surviv'ors' predictions. Journal of Interpersonal Violence, 15(1), 75-90.

140

 

 

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

[n\[wljlj[[[[[[|[[

\
9 7

 

l