ABSTRACT MILK PRODUCTION AND RELATED FACTOR RESPONSE OF DAIRY HERDS AFTER ENROLLMENT ON A PRODUCTION TESTING PROGRAM BY Edward A. Schramski Michigan DHIA Holstein herds of at least 15 cows per year between 1963 and 1967 were the source of data. If the herd was on test perior to 1963 and on test at least three years 1963-1965 it was termed an established herd. A new herd was one that started in 1963, 1964, or 1965 and remained on test at least three years. The study included 1031 established and 348 new herds. The relation of various measures of herd management to herd average production was studied. Sources of variation in herd averages were defined and differences between new and established herds determined. Also changes in herd averages were investigated as time on the testing program continued. The various measures of herd management included the 305 day lactation Mature Equivalent average (M.E.), average number of cows, average age of cows, percent cows Edward A. Schramski in milk, pounds grain, pounds hay, pounds silage, pounds TDN and relation of total to average number of cows. The correlation of the measures excluding the sub- jective factors 305 M.E. and lbs. TDN in the prediction equation with herd average milk or fat production resulted in a coefficient of determination (R2) of 0.40. The two most important factors were lbs. grain and percent cows in milk. A majority of the variation in annual herd produc- tion was due to differences among herds. Variation in hay feeding, average age, and percent cows in milk was more equally divided between yearly variation and among herds. A majority of the variation in herd size was among herds. About half the variation in grain and silage feeding was among herds. County variation explained a larger percent- age of the variation in grain and silage level than in other factors. Established herds produced more milk than new herds on test. Established herds were smaller and older on the average with a little higher percent cows in milk. These herds fed less grain but more hay and silage than new herds on test for three years. Calendar year on test seemed to affect herd average production, 305 M.E. production, grain level and hay feed- ing. New herds followed trends of established herds but were affected more extremely. Herd age seemed to be Edward A. Schramski related to rate of expansion which was about the same for all years except 1967. Changes in level of grain feeding were directly related to production while an inverse rela- tion of level of hay feeding and production was indicated. More variation was present in herd averages of new herds than established. When variation was partitioned among counties, herds within county, and years within herd, established herds were more variable in herd average pro- duction among counties and years within herd. Average herd age, percent cows in milk and silage feeding were also more variable among counties in established herds. With herd average production the dependent varia- ble and the objective factors stated previously as the in- dependent variables the same multiple correlation coeffi- cient (R2) of 0.40 was obtained for new and established herds. In new herds, average age and grain were a little less important. The relation of total—average number of cows was of more importance in new herds. When 305 M.E. was included as an independent variable it was not as im- portant in predicting new herd average production. Production increases for new herds on test were not statistically significant the first three years on test. A positive linear trend was present but with the variation present the improvement was not significant. Edward A..Schramski Average age and herd size changed little due to testing. The percent cows in milk remained constant over the three years on test. TDN level increased during these three years on test due to an increase in silage greater than decreases in grain and hay level. MILK PRODUCTION AND RELATED FACTOR RESPONSE OF DAIRY HERDS AFTER ENROLLMENT ON A PRODUCTION TESTING PROGRAM BY A Edward Afichhramski A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Dairy 1969 ACKNOWLEDGEMEN TS I wish to express sincere appreciation to Dr. John A. Speicher for his assistance during the course of this research as well as his guidance in preparing this manu- script. Also I am grateful to Dr. P. W. Spike for his ad- vice and knowledge of techniques used in this study. I'm very thankful to Mrs. Joanne Landis and es- pecially to Mr. James Flanagan for doing the great amount of computer programming. Also, I thank Mr. A. J. Thelen, Data Processing Supervisor of Dairy Herd Improvement Assocaition Incorpor- ated, for providing technical assistance with the machines and for access to the D.H.I.A. records necessary. ii TABLE OF CONTENTS INTRODUCTION 0 O O O O O O O O O 0 O O O O 0 REVIEW OF LITERATURE . . . . . . . . . . . . Factors Affecting Herd Average Milk and Fat Production . . . . . . . . . . . . Average Number of Cows Average Age of Herd 305 Day Mature Equivalent Milk and Fat Production Percent Cows in Milk Hay, Grain, Silage, and TDN Variations in Herd Averages . . . . . Comparison of Established and New Herds and Changes in New Herds . . . . . . . EXPERIMENTAL PROCEDURE . . . . . . . . . . . Description of Data and Definition of Factors Studied . . . . . . . . . . . Determining Effect of Specified Factors on Herd Production . . . . . . . . . . EXplaining Variation Between Herds . . . Differences Between New and Established Herds Studying Changes in New Herds RESULTS AND DISCUSSION Production as Affected by Specified Factors Explaining Variation of Herd Averages Differences Between New and Established Herds Changes in New Herds CONCLUSIONS iii 21 22 23 23 33 4O 57 60 Page SUMMARY 0 O C O O O O O O O O O O O O O O O O O O O 6 2 LITERATURE CITED 0 O O I I I O O O O O O O O O O O 69 APPENDIX 0 O O O O O O C I O O O O O O O O O O O O 71 iv Table 10 11 LIST OF TABLES Lactation results as affected by calving interval . . . . . . . . . . . . . . . . Changes in 1966-1967 Wisconsin yearly herd averages as related to percent days in milk 0 O O O O I O O O O O O O O O O The effect of grain, silage, hay and pasture on milk production . . . . . . . Estimates of variance components for states, counties, herds, and among years within herds . . . . . . . . . . . Repeatability of herd averages from year to year 0 O O O I O O O O O O O O O O 0 Correlation coefficients and percentage of variation explained by specified factors affecting herd average milk pro- duction . . . . . . . . . . . . . . . . Correlation coefficients and percentage of variation explained by specified factors affecting herd average fat pro- duction . . . . . . . . . . . . . . . . Simple correlations of herd character— istics, all herds combined . . . . . . . Correlation of three measures of produc- tion with each other and specified physical inputs . . . . . . . . . . . . Means and standard deviations of inputs and outputs of established DHIA herds 1963-1967 0 o o o o I o o o o o o o o o Partitioning of established herd varia- tion 1963-1967 0 o o o o o o o o o o o o Page 12 14 15 24 25 31 34 35 37 Table 12 13 14 15 16 17 18 Comparison of established herds and her herds their first, second and third year on test . . . . . Changes yearly 1963-1967 of new and established herd averages Components of variation and partitioning of variation for new and established herds O O O O O O Specified factors plained variation herd average milk Specified factors plained variation herd average milk milk excluded . . Number of complete records as a per- centage of average number of cows in herds as related to year on test Change in herd averages during three years on test with year effect removed accounting for ex- in new and established production accounting for ex- in new and established production, with M.E. vi Page 41 45 49 53 54 54 58 Appendix Table 1 LIST OF APPENDIX TABLES Linear and quadratic, regression coeffi- cients and beta weights of independent variables affecting herd average milk production in established herds when M.E. milk was included . . . . . . . . . Linear and quadratic, regression coeffi— cients and beta weights of independent variables affecting herd average milk production in established herds when M.E. was included and with feed ex- pressed as TDN . . . . . . . . . . . . . Linear and quadratic, regression coeffi- cients and beta weights of independent variables affecting herd average milk production in established herds when M.E. milk was not included . . . . . . . Linear and quadratic, regression coeffi- cients and beta weights of independent variables affecting herd average milk production in established herds when M. E. milk was not included and feeds ex- pressed as TDN . . . . . . . . . . . . . Linear and quadratic, regression coeffi- cients and beta weights of independent variables affecting herd average milk production in new herds when M.E. milk was included . . . . . . . . . . . . . . Linear and quadratic, regression coeffi- cients and beta weights of independent variables affecting herd average milk production in new herds when M.E. milk was included and the feeds expressed as TDN . . . . . . . . . . . . . . . . . . vii Page 71 72 73 74 75 76 Appendix Table 7 Page Linear and quadratic, regression coeffi- cients and beta weights of independent variables affecting herd average milk production in new herds when M.E. milk was not included . . . . . . . . . . . . . 77 Linear and quadratic, regression coeffi— cients and beta weights of independent variables affecting herd average milk production in new herds when M.E. milk was not included and feeds expressed as TDN . . . . . . . . . . . . . . . . . . . 78 viii INTRODUCTION Large variation exists among the levels of milk production in dairy herds. There are differences between tested and non-tested herds. In 1967, production in Mich- igan Dairy Herd Improvement herds was 12,533 pounds com- pared to 9,430 pounds in all herds including tested herds. Herds on a production testing program definitely produce more than non-tested herds. It would be of value to know what factors contribute to the higher production in tested herds and also when the increases in production occur. The objectives of this study were to determine the relation of various measures of herd management to average milk and fat production, to define sources of variation in herd average production, to determine differences between new and established herds, and to investigate changes in herd averages as herds continue on the testing program. The various measures of herd management included the 305 day lactation Mature Equivalent average (M.E.), average number of cows, average age of cows, percent cows in milk, pounds (lbs.) grain, lbs. hay, lbs. silage, lbs. total di- gestible nutrients (TDN), and relation of total number of cows to average number. The source of data for this study was Michigan Dairy Herd Improvement Association records. REVIEW OF LITERATURE Factors Affectinngerd Average Milk and Fat Production Average Number of Cows With increased mechanization one man can handle more cows than before possible. Dairy herds are getting larger and as this happens, there appears to be a negative effect on average production per cow. A number of studies have shown a small negative correlation between herd size and milk production per cow. Speicher (14) found a correlation of -0.05 between number of cows and milk production per cow. In a multiple re— ‘gression analysis, Speicher found number of cows accounted for 3.9% of the measured variation in milk per cow while other variables considered included lbs. grain fed, lbs. hay, percent cows in milk, days on pasture, improvements per cow, cows per man and crop yield index. The total R2 was 0.27. A study by McKinney, Welch and Fosgate (8) showed a nonsignificant negative regression of herd size on fat- corrected milk (FCM). Though statistical significance was lacking the authors stated production per cow tends to de- crease as cow numbers per herd increase. Stone, Burke, Ainslie and Van Vleck in 1966 (17) indicated a slightly negative correlation of herd size and production. Hansen, Barr, and Wieckert (4) found a nega- tive correlation of -0.04 between herd size and milk pro- duction. In 1952 Bayley and Heizer (1) found the standard partial regression coefficient statistically significant for regression of milk and fat yield on herd size. A to- tal decline of 775 pounds of milk was associated with a change in herd size from 20-49 cows, -521 lbs, from 50-79 cows. Only a slight decrease could be predicted when herds contained 80 cows or more. An investigation by Conlin, Corley and Tyler (3) in 1964 found herd size accounted for 1% of monthly and 4% of the 12 month rolling average vari- ation in milk and fat yield. Miller and Dickinson (11) showed a decrease of 1.1 pounds of milk in annual herd average per each additional cow when grain, silage, hay, pasture, and percent days in milk were held constant. The usefulness of herd size in predicting milk production was found to be of little or no value. Average Age of the Herd It is well established that age has a definite ef- fect on level of milk production. Maximum production occurs at six to eight years of age. An increase in pro- duction of 30% is generally assumed from first calving to maturity but for individual cows the increase is frequently much less due to many factors (2). As the cow becomes old- er, there is also a slow but persistent decrease in the fat percentage of the milk. The test usually does not drop more than 0.2 to 0.3 percent during the entire life- time of the cow so is not important from a practical point of view (2). Theoretically the herd average for an older herd should be higher if genetically equal and under the same environmental conditions. But Conlin, Corley, and Tyler (3) could not account for any of the variation between un— adjusted monthly and 12—month rolling averages for milk and fat yield with herd age. 305 Day Mature Equivalent Milk and Fat Production 305 day M.E. production is a measure of production ability and environment. Managers account for a majority of the environmental effect by manipulation of herd inputs and for the genetic effect by selection. The 305 day M.E. is an indication of management of individual cows that com- pleted 305 day records adjusted to a common basis by ex- pressing the records on a mature equivalent basis. In other words it is an indicator of treatment of cows as in- dividuals as opposed to treatment as a herd. The author found no study where 305 M.E. was in— cluded in a prediction equation of herd average milk yield. A high correlation between this factor and milk yield would be anticipated since the production of a por— tion of the herd would be used to predict the production of the entire herd (part-whole relationship). Percent Cows in Milk Speicher and Meadows (16) found that lactation milk production, length of lactation, days Open and days dry increased as calving interval increased. The number of lactations decreased from 4.9 to 3.8 as the calving in- terval increased from less than 366 to more than 425 days. (Table 1) Table l.--Lactation results as affected by calving inter- val Calving Interval Length of Days Days Number of Days lbs. Lactation Open Dry Lactations 365 13031 301 76 53 4.9 366-395 13680 317 100 62 4.6 396-425 14380 338 128 69 4.3 426 15510 381 185 80 3.8 When cows were divided into production groups, calving intervals were shorter for the low production groups. The researchers hypothesized it was due to the fact that (1) lower producing cows encountering conception difficulties are removed from the herd more rapidly than their counterparts with higher production and (2) breeding of higher producing cows is purposely delayed. The first hypothesis is an economically sound practice. The second is not since average daily returns over feed cost for each productive group decreases as calving interval increases. The average cost of delayed conception beyond 85 days after freshening was .50, .78, and .77 cents per day for groups averaging 96,116, and 146 days open per year. Crowley (5) stated days in milk is a general meas- ure of breeding efficiency and a generally accepted ratio is to have cows milking 84% of the time. The normal recom- mendation is that the herd should average 87-90% cows in milk since cows culled while in production have no dry days. If conception is delayed beyond 60-70 days after calving, the dairy cow will produce more milk in a given lactation, Legates and Louca (6). If a cow is open for the entire 305 day lactation period, her production may be increased as much as 10% over that expected with conception at 60 days postpartum. In 1962, Smith and Legates (13) found the relationships between 305 day records and days open may be interpreted to be due largely to the influence of gestation on production. In a study by Louca and Legates (7) it was found days open are not uniformly expensive for all lactations of an animal. A calving interval of 13 months for first calvers and 12 months for second and later calvers was suggested as an optimum length for attaining maximum pro- duction. An increase of 1.16 kg. of milk for each addi- tional day open was obtained for first lactations. For second and third lactations there was a decline of 3.58 and 3.68 kg. of milk, respectively for each additional day open. The expected loss in net income for each additional day Open would be approximately 25-70 cents. Crowley (5) stated yearly herd averages decrease as percent days in milk decreases (Table 2). Speicher (15) found a correlation of 0.26 between percent days in milk and milk production. In the same Table 2.--Changes in 1966-1967 Wisconsin yearly herd aver- ages as related to percent days in milk % days in milk No. of herds Milk (lbs.) 78-82 4,357 11,044 83-87 16,203 12,270 88-92 5,936 12,821 multiple regression analysis cited earlier, percent cows in milk accounted for 46.1% of the variation explained. Stone, Burke, Ainslie, and Van Vleck in 1966 (17) found percent days in milk accounted for 11% of the varia- tion explained in milk production (R2=0.41). Other varia- r1 bles were grain feeding, succulent forage, dry forage, net I energy from pasture and herd size. A study by Hansen, Barr, Wieckert (4) showed a i- correlation of 0.37 between percent days in milk and milk E produced. They also found a standard partial regression of 0.33 on milk yield (P<.01). An investigation by Conlin, Corley, and Tyler in 1964 (3) found percent days in milk accounted for 39% of the monthly and 30% of the lZ—month rolling average varia- tion in milk and fat yields. Percent days in milk and stage of lactation were the major sources of variation in monthly averages whereas stage of lactation appeared to have less influence on the 12—month rolling average. These two factors were studied in addition to herd size and herd age. With all four factors they accounted for 79% of the total variation in monthly milk averages and 43% of the total variation in lZ-month rolling average milk yield. In 1968, Miller and Dickinson (11) found that for each 1% increase in percent days in milk, annual herd 10 average milk increased 115.6 pounds when grain, silage, pasture and herd size were held constant. Hay, Grain, Silage, and TDN Feeding has changed over the years. Nationally, grain feeding has increased 50% per cow in DHIA records in the last 10 years (11). Michigan DHIA farmers reported they fed 3,247 lbs. per cow in 1958 and 4,548 lbs. per cow in 1967, about a 40% increase. There has also been a change from feeding dry hay to feeding more corn silage or hay silage. There has been much work done in this area. In some unpublished work Speicher (15) found unit increases in grain per cow per day were associated with an output curve for milk which increased at a decreasing rate. The correlation between milk produced and pounds of grain per cow was 0.41. Grain accounted for 31.9% of the variation explained and hay ac- counted for 7.3% in a multiple regression (R2=0.27). In a study by McKinney, Welch and Fosgate, 1965 (8) methods of grain feeding and their effect on fat cor- rected milk (FCM) were observed. Those herds that fed ac- cording to production by weighing or measuring the grain produced 392 pounds of FCM over the average of 8,481 lbs. FCM. The group that fed grain by estimating a predeter- mined amount based on production were 110 lbs. above aver- age. These two groups contrast with those that fed a fixed 11 amount (not according to production) who were 510 lbs. FCM below average. Hay showed no significant effect on milk production in this study. In 1966, Stone, Burke, Ainslie, and Van Vleck (17) made a study of New York herds. They reported a trend toward more concentrate and succulent feeds but less dry forage and pasture. An increase of 1 kg. concentrate was found to cause 0.84 kg. more milk if other factors were held constant. Concentrates accounted for 28% of the var- iation explained in milk production when succulent forage, dry forage, net energy from pasture, herd size and days in milk were also considered. They stated changes in milk production are fairly closely associated with changes in succulent or dry forage fed. A study by Miller, McDaniel and Creegan (12) dealt with variance components for amounts of concentrates and hay fed. Of the variance in grain and hay feeding, 6.3 and 4.4% were between states, 8.9 and 7.8% between counties 44.4 and 47.6% between herds, and 40.4 and 40.2% between years within herds, respectively. The biggest variation was between herds and within herds in different years. Quantities of concentrates and hay fed were correlated with milk produced; .05 and .00 totally and .59 and .05 among herds within county. This indicated concentrate levels provide information on a within herd basis but not 12 among counties and states, due to other factors having more effect on production. In 1968, Miller and Dickinson (11) studied the ef- fect of feed on milk production (Table 3). Table 3.--The effect of grain, silage, hay and pasture on milk production Feed Effect on Milk Production Grain (100 lbs.) +105 change in milk production per unit change Silage (100 lbs.) + 9.4 change in milk production per unit change Hay (100 lbs.) + 18.8 change in milk production per unit change Pasture (days) + 5.4 change in milk production per unit change Concentrates were the most useful in predicting milk yield, 27.7 times greater than pasture; silage 3.4 times as useful as pasture; and hay 1.5 times as valuable, based on a comparison of amount of variation explained by the factor asla ratio to the amount explained by days on pasture. There has been some work done with the effect of TDN on milk production. Bayley and Heizer in 1952 (1) found that for an increase of 1 lb. of TDN daily per 1000 lbs. body weight, there was an average increase of 551 13 lbs. in average milk production per cow and 18 lbs. of fat. The study was one year and 8 months in length and feeding measures were based on rations fed during winter months in the barn. A study by Hansen, Barr, and Wieckert (4) found a correlation of 0.14 between kgs. TDN per 455 kgs. body weight and milk produced. Variations in Herd Averages Variation among herd averages were investigated to determine variation due to location, herd environment and within herd effects from year to year. The only previous work found on this subject was by Miller, McDaniel, and Creegan (12). The study involved Northeastern United States herds and applied only to herds feeding hay and grain. Following are some of their esti- mates of variance components for states, counties, herds and among years within herds expressed as percent of total variance (Table 4). County components tended to follow the pattern of state estimates. The herd components have the largest variation. Miller and Dickinson stated part of the reason why herd components of economic measures were high was be- cause dairymen price home grown grain and feeds at esti- mated unit costs of production which would be inflationary. Table 4.--Estimates of variance components for states, counties, herds, and among years within herds Components OZState 02County ozHerd czYear Milk 2.4% 3.3% 61.6% 32.7% Fat % 24.8 9.5 43.9 21.8 Concentrate 6.3 8.9 44.4 40.4 Dry forage 4.4 7.8 47.6 40.2 Cost of concentrate 17.3 11.3 37.0 34.3 Feed cost 23.4 14.3 29.0 33.3 Income over feed cost 4.7 9.6 61.9 23.8 The herd component of variance was found to be most important for milk yield and income over feed cost. This indicated the importance of a dairyman's competence in his management and feeding program. The variance components for years within herd in— dicated changes in herd management, genetic merit of cows, and prices from year to year. The following are repeatabilities that were calcu- lated for certain factors by Miller, McDaniel, and Creegan (12) (see Table 5). Higher variation was present yearly in feeding and feed costs than production. Income over feed cost was highest in repeatability because of a relatively constant Inilk price. 15 Table 5.--Repeatability of herd averages from year to year Measure Repeatability Milk .65 Fat % .67 Concentrate .52 Dry forage .54 Cost of concentrate .52 Feed cost .47 Income over feed cost .72 Comparison of Established and New Herds and Changes in New Herds McKinney, Welch, and Fosgate (8) found the regres- sion of FCM on months on DHIA was not significant. These workers expected it was due to the herds being on test for an average of 66 months and further increases in produc- tion due to culling after two years of testing and culling would be expected to be relatively small. In a study by Stone, Burke, Ainslie, and Van Vleck (17) in 1966 changes in herd averages of 688 herds between 1960—1964 were investigated. Changes of multiples of 227 kilograms of milk and 181 kilograms of concentrate were the bases of grouping. The largest positive changes were Inade by the herd lowest in that factor at the time, and 16 vice versa. In following years there was a slight tendency to reverse the magnitude of the change, but by the third year there was no definite pattern of change. They found a definite tendency for herds to move toward the average. Miller and Meadows (9) undertook a study to deter— mine whether herds continuously on test increased in pro- duction levels over herds just starting to test and non- tested herds in the same period of time, 1962-1964. In 1962 the production differences between non- tested and new herds on test were not significant. The tested herds exceeded these two groups by 1832 lbs. milk. Between 1962 and 1964 those herds that were on test before 1962 increased their production level 1663 lbs. to an average of 14,795 lbs. During this same time interval the herds starting to test in 1962 increased their production level 1428 lbs. to an average of 12,957 making their in- crease almost the same as herds continuously on test. The non—tested herds at the same time increased their produc- tion level 891 lbs. to an average of 11,914 lbs. milk, about one-half as much increase as the tested groups. Miller and Meadows concluded that there are pro- duction differences between tested and non-tested herds which diverge with time. These differences are minimized by starting to test. Also that increases are about the same percentagewise for new herds on test as for the con- tinuously tested herds. EXPERIMENTAL PROCEDURE Description of Data and Definition of Factors Studied Michigan Dairy Herd Improvement Association records were used for analysis. Herds were included in the study if coded as a Holstein herd averaging at least 15 cows for each year and enrolled in the Dairy Herd Improvement Asso- ciation (DHIA) for a minimum of 3 consecutive years be- tween 1963 and 1967. Established herds were those on test before 1963 and remained on test at least through 1965. New herds were those that started testing in 1963, 1964, or 1965 and remained on test at least three years. A total of 1379 herds were included in the study--1031 established herds and 348 new herds. The new herds were composed of 123 that started on DHIA in 1963, 119 in 1964, and 106 herds in 1965. Annual herd summary cards were used to collect or calculate the following data: Mature equivalent milk--Average of completed 305 day lac- tations for milk expressed as mature equivalents. Mature equivalent butterfat--Average of completed 305 day lactations for fat expressed as mature equivalents.~ 17 18 Average age-~Average age of cows at calving expressed in months. Average number of cows—-Total number of cow days divided by 365. Percent cows in milk-—Cow days in milk divided by total Pounds Pounds Pounds Pounds Pounds Pounds cow days. of milk--Herd average milk, total milk produced di- vided by average number of cows. of butterfat-—Herd average fat, total butterfat In. produced divided by average number of cows. of grain--Tota1 pounds of grain fed divided by av- erage number of cows. of hay--Total pounds of hay fed divided by average number of cows. of silage-~Tota1 pounds of silage fed divided by average number of cows. of TDN——=(lbs. grain)(0.75)+(lbs. haY)(0.50)+(lbs. silage)(0.20). Total—average number of cows=Total cows that have been in in herd any portion of the year - average number of cows Total cows that havegbeen in herd any portion of the year. Determining Effect of Specified Factors on Herd Production Least-squares multiple regression equations were solved to determine the relationship of M.E. milk, M.E. 19 fat, average age, average number of cows, percent cows in milk, pounds of grain, pounds of hay, pounds of silage, lbs. TDN, and total-average number of cows with herd av- erage production of milk and fat. Means and simple corre- lations were also computed at this time. The data were included from all selected herds. Two main sets of multiple regression functions were solved for pounds of milk and pounds of fat as the. respective dependent variables. Four sets of independent variables and their squares were used: 1) M.E. lactation average milk or fat (depending on which variable), average age, average number of cows, percent cows in milk, pounds grain, pounds hay, pounds silage and total-average number of cows. 2) Deletion of M.E. lactation average and retaining the other independent variables. 3) M.E. lactation average, average age, average number of cows, percent cows in milk, lbs. TDN and total- average number of cows. 4) Deletion of M.E. lactation average and solving the regression with the remaining independent variables of #3 above. The least—squares program gave beginning estimates with all independent variables included and then deleted I'll 14- -‘fi-‘ V I1 20 least important variables singularly until each remaining variable met the requirement of a 0.05 significance level. To estimate the importance of the independent var— iables on herd production, their influence through direct and indirect effects were calculated. For a linear func- tion or quadratic function, the variation in herd average accounted for by the factor was the square of the beta weights or'standard partials. Direct and indirect effects fx- y=?-.—+2 +2 . rXiXiz 2. o 1 on 8 X1 8 X12 (BXIBXiZ ) where 8 X1 and Bzxi2 were considered to be the direct effects of Xi and xi2 (the linear and quadratic effect of each indepen- dent variable) on Y, and the covariance (product of the beta weights times the simple correlation coefficient be- tween xi and Xiz) multiplied by 2 as the indirect effect of the factors through each other on herd average produc- tion (18). The sum of the path coefficients of all variables was divided into individual variable effects to find the percentage of variation each independent variable con- tributed to explaining herd variation in production. Only variables significant at the 0.05 level in linear or quad- ratic effect were used to calculate the percent variation of the dependent variable each explained. 21 Explaining Variation Between Herds Analyses of variance were used to test variation between herds. Established and new herds were computed separately. Components of variance were computed to find variation between counties, between herds within counties, and between years within herd. Each of these variance components divided by the total distributed the variation among the three. Differences Between New and Established Herds Three methods were used to compare differences be- tween new and established herds. Multiple regression equations were calculated for established herds and new herds separately. This was done using the same variables and sets of variables as previously when all herds were together. Path coefficients were computed for the two groups when: (l) M.E. milk lactation average, average age, average number of cows, percent cows in milk, pounds of grain, pounds of hay, pounds of silage and total-aver- age number of cows were independent variables; (2) the same variables with M.E. milk lactation average left out. The two groups were compared using the path coef— ficients to determine if certain factors explained more of the variation between the two groups. 22 Secondly, an analysis of variance was utilized to compare variation differences and allocation between es- tablished and new herds. Means of the two groups were studied to look for differences and trends in averages. Simple correlations were used to compare relation of factors between the two groups. Studying Changes in New Herds Averages of M.E. milk, M.E. fat, average age, av- erage number of cows, percent cows in milk, pounds of grain, pounds of hay, and pounds of silage were studied the first three years on test. Total-average was omitted because it was not significant in previous calculations. To determine the effect testing had on new herds, established herds were designated to reflect year effect. An overall average for each of the independent variables and milk production was calculated for established herds. This was subtracted from established herd averages for each of the years 1963-1967. The results were adjustment factors used on new herds. The year effect was added to each of the new individual herd averages. Then a test for non-linearity was solved with Y, herd average milk produc- tion and X, the year on test. RESULTS AND DISCUSSION Production as Affected by Specified Factors The regression of specified factors on milk or fat production resulted in correlation coefficients of 0.86 to 0.87 when a measure of mature equivalent milk or fat was included. Exclusion of these measures of production as independent variables reduced the correlation coefficients to 0.33 and 0.40. Table 6 shows the correlation coefficients realized with various combinations of factors as independent varia- bles and herd average milk as the dependent variable. Also included in the table is the percentage of explained var- iation each factor accounted for as computed by path co- efficients. Similar statistics are shown in Table 7 for the regression of the specified factors on herd average fat production. Multiple regressions of each set of independent variables on milk and fat production showed essentially no difference in correlation coefficients. Referring to Tables 6 and 7 it is observed that the study's most accu- rate independent variable combination designated GHS re- sulted in correlation coefficients of 0.40 and 0.39 for 23 Table 6.--Correlation coefficients and percentage of vari- ation explained by specified factors affecting herd average milk production M.E. GHsl Factors GHS2 M.E. TDN3 TDN4 M.E. milk 96.00%5 ---- 96.85% ---- Ave. age 0.52 0.60% 0.44 0.87% Ave. no. cows 0.17 4.31 0.23 10.03 % cows in milk 2.36 35.11 2.15 44.27 Lbs. grain 0.74 56.02 ---- ---- Lbs. hay 0.03 2.40 ---- ---- Lbs. silage 0.00 1.57 ---- —--- Lbs. TDN --~- ---- 0.24 44.51 Tot—ave no. cows 0.07 0.00 0.08 0.32 100% 100% 100% 100% R2= 0.87 0.40 0.86 0.33 hay and silage as separate variables. TDN . lM.E. milk as an independent variable with grain, 2M.E. milk omitted from the independent variables with grain, hay, silage as separate variables. 3M.E. milk an independent variable with grain, hay and silage combined and expressed as TDN. 4M.E. milk excluded from the independent variables and grain, hay and silage are combined to be expressed as 5The percentage of explained variation (R2) that this factor accounted for as computed by path coefficients. 25 Table 7.--Correlation coefficients and percentage of vari- ation explained by specified factors affecting herd average fat production Factors M.E. GHSl GHs2 M.E. TDN3 TDN4 M.E. fat 96.57%5 -—-- 96.98% ---- Ave. age 0.52 0.65% 0.46 0.92% Ave. no. cows 0.16 5.73 0.21 9.79 % cows in milk 2.18 37.08 2.11 46.34 Lbs. grain 0.52 51.55 ---- ---— Lbs. hay 0.02 1.69 ---- -—-- Lbs. silage 0.00 3.31 ---- ---- Lbs. TDN -—-- ---— 0.20 42.95 Tot ave no. cows 0.03 0.00 0.04 0.00 100% 100% 100% 100% R2= .88 .39 .87 .33 hay and silage as separate variables. with grain, hay and silage as separate variables. hay and silage combined and expressed as lbs. TDN. lM.E. fat as an independent variable with grain, 2M.E. fat excluded from the independent variables 3M.E. fat is an independent variable with grain, 4M.E. fat is omitted from the independent variables and TDN remains as a combination of grain, hay and silage. 5The percentage of explained variation (R2) that this factor accounted for as computed by path coefficients. 26 milk and fat, respectively. That designated TDN resulted in a correlation coefficient of 0.33 for both milk and fat. In the two remaining sets of independent variables where appropriate M.E. measures were included, the coefficients determined were 0.87 versus 0.88 and 0.86 versus 0.87 for milk and fat, respectively. In each instance a combination of the same or comparable independent variables resulted in correlation coefficients within 0.01 of each other. A comparison of percent variation explained by each factor as determined by path coefficients substanti- ates the similarity in results. Small differences existed in the independent variable combination GHS between milk and fat. The amount of variation explained by age is simi- lar. Herd size accounted for about 1% more of the varia- tion in fat production than in milk. Since average number of cows did not explain much of the variation in production for each of them, this difference is not important. Per- cent cows in milk accounted for 35.11 and 37.08% of the explained variation in milk and fat yield, respectively. This is relatively close, no important difference was present here. Pounds grain has the largest difference: 4.5% more of the explained variation in milk yield is ac— counted for by grain than in fat yield. It is not known if this difference is significant but 51.6% and 56% are similar enough to warrant discussing the effect of grain on milk and fat as one. Since hay and silage accounted 27 for little of the variation, no significant differences exist between these factors in explaining variation in milk and fat yield. In the combination of independent variables desig- nated TDN, only small differences are present between the variables explaining variation in milk and fat production. Percent cows in milk explains 2% more of the fat variation than that of milk. Pounds TDN accounts for about 1.5% more of the variation in milk yield. Again small differ- ences are present but not large enough to warrant discus— sion separately. In the two sets of independent variables where ap- propriate measures of mature equivalents were included, the results varied only by a fraction of a percent, making the factors extremely similar in accounting for explained variation in milk and fat yield. These results justify discussing the results with milk production; implying fat is affected the same way. With milk production the dependent variable and the objective factors average age, average number of cows, percent cows in milk, pounds grain, pounds hay, pounds silage and total-average number of cows as independent variables the two most important inputs were pounds grain and percent cows in milk. These tWo factors accounted for 91% of the explained variation which is (0.4 X 0.19) 36% of the total variation in herd average milk production. 28 Grain accounted for 56% of the explained variation compared to approximately 36% by percent cows in milk. The simple correlations of grain and percent cows in milk with herd production are 0.49 and 0.43 respectively. When mature equivalent was included as an indepen- dent variable, percent cows in milk accounted for more of the explained variation than grain did. M.E. milk absorbed some of the effect of all the other independent variables. It is probable that more of grain's effect was absorbed than percent cows in milk because of the higher correlation of grain and M.E. milk. The importance of feeding and breeding efficiency is very clear. Farmers must feed as much grain as econom- ically feasible to get high production. Along with this cows must be bred back as soon as possible after about 60 days so a high percentage of the cows are contributing to herd production. Next in importance explaining variation was the average number of cows in the herd. Herd size accounted for 5% of the explained variation. When the beta weights are examined, milk production decreases to a certain point as herd size gets larger and then starts to increase slowly. It may be milk production goes down as one man handles more cows because not as much individual care could be given to a larger herd. After a certain point the larger herds become a two-man operation where the cows 29 can receive better management and production goes up slowly. Hay and silage accounted for a small part of the explained variation. Although necessary for the cow to live, these feeds do not appear to be as important as .grain affecting production. It should be remembered the weights of hay and silage particularly are estimates by the farmer. Since the weights are estimates part of this small effect may be due to inaccuracy of the data. Also roughage quality is not accounted for and this is a very important factor. As studied here, there is essentially no difference between hay and silage affecting production and either one can be fed depending on land type, degree of mechanization and economic considerations. When grain, hay and silage were combined and ex— pressed as TDN, the feeds account for less of the varia- tion. Here also the accuracy of the assigned TDN values must be considered. The three feeds are estimated by the farmer originally and then percentage of TDN assigned to each were estimates, thereby adding to the possible inac- curacy. Using the values derived, it was apparent that some of grain's effect was absorbed by number of cows and percent cows in milk, for TDN accounts for 45% of the ex- plained variation compared to 60% when the three feeds are separate. It is very possible this resulted because of 30 inaccurate TDN values being assigned, especially to the roughage which is more variable in quality. The relation of total to average number of cows had virtually no relation with herd average production (Table 8). This factor was intended to measure cow turn— over as an indication of culling level. As observed, the relationship was small indicating culling of cows may have been carried out because of reasons other than low produc- tion. Average age of the herd also accounted for little of the variation in milk production. The range of herd age was 3 years 7 months to 5 years and 3 months which in- cluded 95% of the herds. It is possible this variation was not enough to show a difference between young and old herds or that other factors hid the effect of age. When M.E. milk was included as an independent var- iable it accounted for almost all of the variation. How the farmer manages individual cows that complete 305 day records, adjusted for age differences is a good indication of what his herd production will be for the whole year. M.E. milk has a correlation of 0.92 with herd average milk and the majority of cows in a herd do complete their lac- tations thereby resulting in this close relationship. The correlations of actual milk, mature equivalent milk and herd average production were all near 1.00. Also the correlations between the physical inputs and these 31 .cofluoscoum mmmnm>m cummm .mc0flumuomH mop mom ompmamfioo mo mmmum>¢ oo.H moo. bum. moo. pom .m.z oo.H omo. ooo. sane .m.2 oo.H boo. Homo .uoa oo.H axafle .664 monumfi .4.o a l m o m * .cwo # m s z o o m .o.¢ .<.« .m.z .z.z .m.4 .z.« numuomumgo owcHQEoo mono: Ham .moaumoumuomumao who: mo macaumamuuoo mHmEHmII.m magma 32 oo.H Had. ooo. oHo.- oma. oma. mod. omm. mHo. ooa.- mos. omH. moo. mos. .m>m-poe oo.H mom. oom. moo. oom. Hum. Nos. ooa. HNH.- who. mom. omm. mom. zoo .moq mofiumfl .4.B .H. * m t m * .GHU # .m # 2 # U w .UJN .¢.< .m.2 .2.Z .me .ZJN. Inwfiomumno omscfluGOUIu.m magma 33 three measures of production were nearly the same (see Table 9). The only difference was that percent cows in milk was correlated with herd average milk production more than with the other two. This would be possible since the herd average is calculated as total milk produced divided by average number of cows, and if a lower percentage of the cows are in milk they will still be in the divisor result- ing in a lower average per cow. The greatest value is placed upon the regression Y=f(X2,X3,X4,X5,X6,X7), where Y=herd average milk or fat production and X2=average age, X3=average number of cows, X4= percent cows in milk, X5= pounds grain, X6= pounds hay and X7= pounds silage. This gives an R2 of 0.4 which is in line with previous research (4), (10), (11), and (17). These are all environmental effects involving feeding and management. All are objective, all physical inputs are measured and no assumptions have been made as with TDN and no measures of production are present as independent var— iables. If methods were available to measure the genetic ability of these herds more of the variation could be ex- plained. Explaining_Variation of Herd Averages The means and standard deviations of the 1031 es- tablished Michigan Holstein herds included in the study 34 .m3oo wo nomads mmwnm>m o» HMUOBH.¢.B xxHflE pcmam>wsvw mudpmzu.z.z uxHHE Hmsuofiu.z. In: mm.mm III ww.m III «mm In: m3oo mo Hmnfisz mm.ov mo.om mm.mm om.mw m~.m mm.w vm so mom mmmum>4 HN.HN m~.Hm ~m.mm vv.mm sm.m om.oa mmvv hamm pom .m.z Hm.om vm.am mv.on mm.mm mm.m nm.m nvsmmmm Hnmmoov xHHE .m.z no.mm mm.ma mm.mm ma.m> Ho.oa m~.m omow Nmmv umm .m>m chow www.mm wmn.ma wma.mm wvm.vn wmm.m mmm.m mwommmm mmmmmmm xHHE .w>m whom hmmumm 3oz .memm 362 .ampmm 3oz .nmumm 3oz Hmmw Nb tum: No (wwcsoo No mocMflHm> muouomm mo mucmcomaou Hobos mono: tmnmwa Inmumm tam 3w: Mom GOHHMHHM> mo mascofiuflunmm tam cowumflum> mo mucmcomEOUIl.vH magma 50 equivalent production, the variation was allocated simi- larly among the three areas for new and established herds. Differences were present between the two groups in distribution of variation of herd average production. New herds had a greater percentage due to yearly and county variation. When herds first came on test there was much var— iation among them due to management, primarily. The lower herds increase production faster than the better herds, thereby causing less variation as seen in a lower varia- tion among herds in the established group. The higher variation among counties of established herds could be due to the high number of these herds, a1- 1owing wider distribution in more diverse conditions. Es- tablished herds were more variable yearly than expected. Variation among herds decreases after herds are on test but yearly variation does not decrease indicating it is not a controllable factor. New and established herds were similar when vari— ation in 305 M.E. production was allocated among the three areas but differences were present partitioning variation in herd average production. This may be because percent cows in milk would have little affect on 305 day production but would affect herd average production and new herds have more variation in percent cows in milk. 51 Considering the variation present in age, new herds had a lower percentage of the differences among' herds and a higher percentage due to yearly variation. Since new herds expanded at a faster rate they would have more in common saving heifers for replacements and keeping some of the older cows. Established herds decreased in age each year at a relatively constant rate while new herds increased at different rates thereby accounting for less total yearly variation. A greater percentage of the variation in percent cows in milk was present among herds for new herds than established. This means more variation is present among the managers of new herds on test and their abilities to keep a high percentage of cows in milk and these differ- ences decrease as testing goes on. Yearly variation within herds did not decrease as fast, with testing, for yearly differences explained more of the total variation in estab- lished herds than variation among herds. Differences in variation of grain feeding were small between new and established herds. Managers of new herds varied more in the amount of grain fed. In these new herds though a smaller percentage was due to yearly variation. As on tests, differences between managers de- crease faster than does yearly variation which is affected by weather conditions, grain price, milk price and other factors. 52 The two groups had about the same variation in hay feeding but established herds had a greater part of the variation due to differences between years. New herds were more variable in the level of silage fed. A large county difference was present between new and established herds, with established herds being more variable. New herds may have come from the same general locality that did not differ as much in the feasibility of growing corn. Managers of new herds on test were more variable in silage feeding; the higher variability in size of new herd may be related. In addition, a greater percentage of variation was due to yearly changes which may have been caused by the faster increase in silage feeding of new herds. A comparison was made to determine if certain fac- tors accounted for more of the explained variation in new herds than established herds. Tables 15 and 16 show the comparison. Table 15 includes M.E. milk as an independent variable while Table 16 just included the objective factors With herd average milk the dependent variable and average age, average number of cows, percent cows in milk, pounds grain, pounds hay, pounds silage and the relation of total-average number of cows as independent variables, both new and established herds resulted in a multiple cor- relation coefficient of 0.40. 53 Table 15.--Specified factors accounting for explained var- iation in new and established herd average milk production Factors New Estab. M.E. milk 54.12%1 96.84% Average age 3.88 0.62 Average number of cows 3.16 0.14 Percent cows in milk 26.45 1.89 Grain (lbs.) 11.30 0.46 Hay (lbs.) 0.00 0.02 Silage (lbs.) 1.09 0.00 Total-average number of cows 0.00 0.02 100% 100% R2= 0.85 0.87 lThepercentage of explained variation in herd production the factor accounted for as computed by path coefficients. When M.E. milk was included as an independent var- iable, new herds had a correlation coefficient of 0.85 com- pared to 0.87 in established herds.' Table 15 shows that M.E. milk accounts for less of the explained variation in herd average production of new herds. Most of this could be due to the fact that the first year on test, the herds have fewer completed records and consequently the accuracy of predicting herd production would be lower. Table 17 54 Table 16.--Specified factors accounting for explained var- iation in new and established herd average milk production, with M.E. milk exc1uded Factors New Estab. Average age --- 0.45% Average number of cows 5.43%1 6.44 Percent cows in milk 32.13 30.55 Grain (lbs.) 56.20 59.32 Hay (lbs.) 3.00 1.58 Silage (lbs.) 1.78 1.65 Total—average number of cows 1.46 0.00 R2= 0.40 0.40 1The percentage of explained variation in herd production the factor accounted for as computed by path coefficients. Table l7.--Number of complete records as a percentage of average number of cows in herds as related to year on test Herds Year on Test lst 2nd 3rd 5th New herds in 1963 0.60 0.80 0.81 0.83 New herds in 1964 0.54 0 83 0.77 New herds in 1965 0.48 55 shows the relation of number of completed records to aver- age number of cows. Since 305 M.E. accounts for less of the variation, the other factors pick up more of the explained variation in order of their importance. When M.E. milk is included percent cows in milk accounts for more of the explained variation than grain. M.E. milk absorbs more of grain's effect than percent cows in milk because of their higher correlation. When 305 M.E. was not included as an inde- pendent variable in Table 16, the two groups were quite similar on how the factors account for the explained var- iation. This is another argument that the differences that occur in Table 15 were due to the lower percentage of completed records the first year on test. Average age of the herd had no significant effect on new herds and only accounted for a fraction of one per- cent of the explained variation. Only a 1% difference existed between new and estab- lished herds on the effect of average number of cows on milk production. The herd size of established herds ac- counted for 1% more of the explained variation. This was not a significant difference. Percent cows in milk of new herds accounted for 1.58% more of the explained variation in herd averages. New herds had a lower percent cows in milk and more varia- tion in percent cows in milk between herds. 56 Grain feeding in established herds accounted for 4% more of the variation explained than in new herds. Grain feeding had more of an effect on herd production in established herds. Established herds got more milk from less grain than did new herds which may be because of bet- ter cows or better management as stated previously. Hay was slightly more important in explaining var- iation in new herds than in established herds. No differ- ence was present between the two groups on how silage af- fected herd production. Established herds were not affected significantly by the indication of culling level, total-average number of cows. In new herds it accounted for 1.5% of the ex- plained variation. This small effect may have been due to a lower culling level of expanding new herds. A function for explaining variation in new herds and one that is comparable to the one suggested earlier for established herds would be Y = f(X3,X4,X5,X6,X7,X8). Where Y = herd average production and X3 = average number of cows, X4 = percent cows in milk, X5 = pounds of grain, X6 = pounds hay, X7 = pounds silage, X8 = relation of total-average number of cows. The relation of total-aver— age number of cows must be included because it explains some of the variation and is an objective factor calculated from herd data. Average age had no significant effect on 57 new herd production so can be eliminated from the predic- tion equation. Changes in New Herds Changes that occurred in new herds on test during their first three years were studied. The first, second and third year on test for the three groups were averaged with calendar year effect taken out. Established herds were assumed to be without initial effects of coming on test and changes yearly represented calendar year effects. It was assumed established herds reacted to year effect in the same manner as new herds. No genetic improvement was assumed during this time period when adjusting for year effect. Table 18 shows the effect of three years on test on specified herd averages adjusted for calendar year ef- fects. A positive linear trend was present in milk produc- tion but variation between herds was such that the improve- ment was not statistically significant. This is comparable with previous work done by McKinney, Welch and Fosgate (8). Most improvement was made the second year on test. The overall regression line had a bl value equal to 48 pounds for each of the three years on test. The herds that came on test were quite high in production when starting. It may be primarily good herds that come on test. These herds could have been on owner-sampler or were doing a good job 58 Table 18.--Change in herd averages during three years on test with year effect removed Factors Year on Test lst yr. 2nd yr. *3rd yr. Herd average milk 12733 12821 12826 Average age 50.97 51.63 52.10 Average number of cows 44 45 45 Percent cows in milk 86.20 85.90 85.97 Pounds grain 4641 4582 4550 Pounds hay 4859 4821 4783 Pounds silage 8957 9442 9887 before deciding to start testing under the standard DHIA plan. Great variation was present between new herds in herd average production. Farmers differ in ability to make use of records so some herds may make excellent progress after coming on test while others make nearly none at all and the average reflects all situations. During this same period of time new herds didn't increase grain as fast as did established herds so after removing year effect the amount of grain fed appeared to have decreased as on test. The advantage of using records for culling on production was most likely negated by feed- ing less grain. 59 The average age of herds increased by one month over the three years on test. If this exerted any effect it would be to increase herd production since fewer young- er cattle would be present. When year effect was removed new herds increased herd size by one cow faster than established herds the first year on test, but expanded at the same rate the sec— ond and third years. Testing had little influence on new herds to increase herd size faster than established herds. Breeding efficiency and/or culling intensity re- mained at the same level the first Three years on test. This indicated new herds on test did as well as established herds. TDN level increased as the herds were on test but this came about because of increases in corn silage great- er than decreases in grain and hay. Grain level decreased about 90 pounds per cow during the first three years on test. This is assumed to have happened due to information received from testing. Evidently farmers did not consider it profitable to feed at the original level or higher. Hay feeding decreased by about the same amount each year while at the same time silage feeding increased 485 and 445 pounds the first and second year on test after year effects were taken out. There was a definite trend to a higher silage feeding program in the new herds. CONCLUSIONS The majority of variation in herd average produc- tion was among herds within county indicating the impor- tance of the manager in influencing herd production. Of the inputs studied, grain feeding and percent cows in milk had the greatest effect on milk production. Average age and herd size have little influence on herd production in normal farm situations. Little difference was present be- tween hay and silage affecting herd production so either can be fed depending on the farmers land type and economic considerations. Also as measured here the indication of culling level had little affect on production but factors other than culling for production are assumed to have caused higher cow turn over. For the most part, new herds on test are affected by the inputs in the same manner as established herds. Average age of the herd was not as important in affecting production in new herds.- The rate of cow turnover was of more importance in influencing herd average production of new herds. This indicated that the rate of cow turnover was more closely related to culling for production pur- poses in new herds than in established herds. Grain did 60 61 not account for as much of the explained variation in herd average production of new herds. It is assumed this is because established herds produced more milk with less grain which could be due to cattle of higher genetic po- tential, and/or a higher level of management. Farmers should recognize that year effect influ- ences the feeding program and milk production. Some of the yearly variation in milk production may be due to grain feeding since changes were directly related. If a bad crop year occurs, farmers should purchase feed to meet requirements instead of feeding less which will result in less production. Other year effects beyond the farmers control exist and should be recognized. Production generally increases after coming on test but slowly. Because large variation is present among herds, it appears farmers differ in ability to use records to increase production. After coming on test a majority of the increase in production depends on the farmers manage- ment ability and some on yearly effects beyond his control. Testing has little affect on herd age, herd size and percent cows in milk. Changes that occur are primari- ly year effects of all herds. New herds increased TDN level due to an increase in silage feeding greater than apparent decreases in grain and hay level during the first three years on test after removing year effect. SUMMARY The objectives of the study were to determine the relation of various measures of herd management to average milk and fat production, to define sources of variation in herd average production, to determine differences between new and established herds, and to investigate changes in herd averages as herds continue on the testing program. The various measures of herd management included the 305 day lactation Mature Equivalent average (M.E.), average number of cows, average age, percent cows in milk, pounds grain, pounds hay, pounds silage, pounds TDN and relation of total to average number of cows. A regression equation with herd average milk or fat as the dependent variable and average age, average number of cows, percent cows in milk, lbs. grain, lbs. hay, lbs. silage and total-average number of cows as inde- pendent variables resulted in the same R2 value, 0.40. The two most important factors were grain and percent cows in milk which accounted for 56 and 35% of the explained variation respectively. Herd size was next in importance accounting for 4% of the explained variation. Average age, hay and silage accounted for 0.60%, 2.4% and 1.6% 62 63 respectively. The relation of total-average number of cows did not account for a significant amount of the ex— plained variation in herd production. When 305 day M.E. was included as an independent variable, it accounted for a majority of the variation and raised the R2 to 0.87 because of the part-whole relation- ship. Converting the three feeds to lbs. TDN resulted in less accuracy indicating the assigned TDN values probably were not accurate. Greatest value was placed on the regression equa- tion Y = f(X2,X3,X4,X5,X6,X7), where Y = herd average milk or fat, X2 = average age, X3 = average number of cows, X4 = percent cows in milk, x5 = lbs. grain, X6 = lbs. hay, and x7 = lbs. silage. These were all environmental effects involving feeding and management. All were objective, measuring physical inputs where no assumptions had been as with TDN and no measures of production were present as independent variables. The average Holstein herd of at least 15 cows and termed an established DHIA herd produced 13211 pounds of milk and 481 pounds of fat. Mature equivalent production of these herds was 14284 and 519 lbs. of milk and fat re- spectively. These herds average 44 cows which were 53 months old on the average. Percent cows in milk was at the recommended rate 86.8. Feeding consisted of 4503, 64 4851, and 9517 pounds of grain, hay, and silage respec- tively. When the variation present in established herds, was partitioned among counties, herds within county and years within herd, a majority of the variation was among herds, followed by yearly and county variation. Production variation was allocated 70% among herds, 20% among years and 10% among counties. Average herd age, percent cows in milk and hay feeding were com- parable in that yearly variation was almost as much or more than variation among herds. A majority of the varia- tion in herd size was between herds, 83% with yearly var- iation very low at 7%. Grain and silage were the factors that had the largest variation due to differences among counties. Var- iation in grain feeding was allocated 56% between herds, 27% among years within herd and 17% between counties. The variation of silage feeding was partitioned similarly. New herds produced less than established herds but approached their level after starting to test. On the av- erage new herds had younger cattle the first three years on test but were on a trend of increasing age. New herds had larger herds than average and expanded at a faster rate. The breeding efficiency and/or culling intensity of the new herds remained about the same and about 1% lower than established herds. 65 The established herds fed less grain than new herds which seem to be on a downward trend also. Estab- lished herds had higher production even though less grain was fed. New herds fed more silage than established herds after three years on test which may be due to their larger herds. Level of hay feeding seemed to be on a downward trend in new herds which also fed a little less their first three years on test compared to established herds. The calendar year on test seemed to affect some factors. Herd average production was affected more ad- versely than 305 M.E. because all the cows in the herd are included. Changes in grain level were directly related to production changes and some inverse relation with hay feed- ing was present. No definite relation could be found be- tween changes in the other factors and production. The average age of established herds was not af— fected by calendar year but decreased at nearly a constant rate each year. New herds varied yearly but no definite pattern could be detected. The herd age did decrease some or increase slower when expansion occurred at a faster rate. The rate of expansion was quite similar for all groups until 1967 at which time all herds expanded more than in previous years. Part of this was assumed to have happened because of the milk price increase in 1966. Percent cows in milk varied by only a fraction of one percent each year with no definite pattern. 66 Grain level was variable year to year and but for one case, all groups changed in grain feeding the same way each year. Since 1964, the grain level increased slower and then started decreasing in 1966. A definite pattern of increasing and decreasing hay level alternate years was present, unlike silage which increased for all groups. More total variation was present among new herds. When the variation of each group was partitioned though, established herds had more variation among counties and years for some of the factors. Variation in M.E. production was explained simi- larly for both groups, but differences were present ex- plaining variation of herd average production. New herds had a greater percentage due to variation among herds but established herds had more variation yearly and among counties. Of the variation present in age, new herds had a lower percentage of the differences between herds and a higher percentage due to yearly variation. A greater percentage of the variation in percent cows in milk was present between herds of new herds on test than established, but the established herds had more variation yearly. Differences in grain feeding between the two groups were small. Variation among herds of new herds on 67 test was greater than established herds and caused a lower percentage due to yearly variation. About the same variation was present in hay feeding with established herds having a greater part due to yearly variation. New herds were more variable in silage level among herds and years within herd, while established herds had more county variation. When the effects of specified factors on herd pro- duction of the two groups were compared few differences were present. With milk production the dependent variable and independent variables average age, average number of cows, percent cows in milk, lbs. grain, lbs. hay, lbs. silage and total-average number of cows, identical R2 values of 0.40 were obtained. Average age was not as im- portant explaining variation of new herds but the relation of total-average number of cows was of more importance than in established herds. Also grain accounted for less of the explained variation in new herds. Average 305 M.E. did not account for as much of the explained variation in new herds and resulted in an R2 of 0.85 compared to 0.87 in established herds when includ- ed as an independent variable. New herds increased production after coming on test but due to the amount variation among herds it was not statistically significant. It is hypothesized that pri- marily good herds came on test that had good management or 68 these herds were on some testing program previous to DHIA enrollment. Testing had little effect on herd age, herd size and percent cows in milk. Changes that occurred were pri- marily year effects of all herds. TDN level increased as herds continued testing due to an increase in silage feed- ing greater than decreases in grain and hay level during the first three years on test. LITERATURE CITED LITERATURE CITED Bayley, N.D., and Heizer, E. E. Herd Data Measures of The Effect of Certain Environmental Influences on Dairy Cattle Production. J. Dairy Sci. 35:540. 1952. Brody, S. Growth and Development. X. The Relation Between the Course of Growth and the Course of Senescence With Special Reference to Age Changes in Milk Secretion. Mo. Agr. Exp. Sta. Res. Bul. 105. 1927. Conlin, B. J., Corley, E. L., and Tyler, W. J. Sources of Variation in Monthly and Twelve-Month Rol- ling D.H.I.A. Herd Average Milk and Fat Yields. J. Dairy Sci. 47:701. (Abstract) 1964. Hansen, L. R., Barr, G. R., and Wieckert, D. A. En- vironmental Influences on Production in 100 Dairy Herds. J. Dairy Sci. 51:1229. 1968. Crowley, J. W. Do Cows Need A Dry Period? Hoard's Dairymen. 113:65. 1968. Legates, J. E. and Louca, A. Days Open are Expensive. Hoard's Dairymen. 113:787. 1968. Louca, A., and Legates, J. E. Production Losses in Dairy Cattle Due to Days Open. J. Dairy Sci. 51:573. 1968. McKinney, W. H., Welch Jr. H. K., and Fosgate, O. J. Estimations of Certain Environmental Influences on Milk Production Based Upon Dairy Herd Improvement Data. J. Dairy Sci. 45:361. 1965. Miller, C. C. and Meadows, C. E. Production Increases of Continuous Tested Michigan Herds Compared to In- creases in Non-Tested Herds. (1962-1964) Presented at American Dairy Science Association Meeting, June 21-23. 1965, University of Kentucky; Lexington, Ken- tucky. 69 10. ll. 12. 13. 14. 15. 16. 17. 18. 70 Miller, R. H. Dairy Herd Improvement Yearly Herd Averages II.’ Predicting Income Over Feed Cost. Animal Husbandry Research Division, U.S.D.A. Belts- ville, Maryland. J. Dairy Sci. 51:1840. 1968.- Miller, R. H. and Dickinson, F. M. Factors Influenc- ing Average Milk Production and Income Over Feed Costs in D.H.I.A. Herds. Dairy Herd Improvement Let- ter ARS-44-205. Vol. 44, No. 4, 1968. Miller, R. H., McDaniel, B. J., and Creegan, M. E. Diary Herd Improvement Association Yearly Herd Aver- ages I. Sources of Variation and Relations Among Measurements. Animal Husbandry Research Division, U.S.D.A. Beltsville, Maryland. J. Dairy Sci. 51:1659. 1968. Smith, J. W., and Legates, J. E. Relation of Days Open and Days Dry to Lactation Milk and Fat Yields. J. Dairy Sci. 45:1192. 1962. Speicher, J. A. Relationship of Dairy Farm Net In- come to Specified Farm Management Factors. Ph.D. thesis, Michigan State University, East Lansing, Michigan. 1963. Speicher, J. A. Unpublished data. 1963. Speicher, J. A. and Meadows, C. E. Milk Production and Costs Assocaited With Length of Calving Interval. Holstein Cows. Michigan State University,-East Lansing. Paper presented at the 62nd Annual Meeting of the American Dairy Association, June 26, 1967. Cornell University; Ithaca, New York. Stone, J. B., Burke, J. D., Ainslie, H. R. and Van Vleck, L. D. Changes in Milk Production Related to Changes in Feeding and Management Practices in D.H.I. A. Herds. J. Dairy Sci. 49:277. 1966. Wright, 8. Correlation and Causation. J. Agr. Res., 20:557. 1921. APPENDIX 71 Appendix Table 1 Linear and quadratic, regression coefficients and beta weights of independent variables affecting herd average milk production in established herds when M.E. milk was included1 , Regression Std. Errors Beta Std. Errors Coefficients of Wts. of Betas Coefficients M.E. milk 0.697 0.057 0.736 0.060 Ave. age 100.964 12.313 0.433 0.053 Ave. no. cows -3.500 1.011 -0.051 0.015 % cows in milk 415.921 90.929 0.726 0.159 Pounds grain 0.266 0.051 0.163 0.031 pounds hay 0.041 0.016 0.042 0.016 Pounds silage 0.008 0.006 0.024 0.018 Tot-ave no. of cows -40.026 205.382 -0.002 0.010 M.E. milk2 0.041 0.020 0.122 0.059 Ave. age2 -0.799 0.111 —0.378 0.053 Ave. no. cows2 0.007 0.005 0.017 0.014 % cows in milk2 -2.033 0.531 -0.608 0.159 Pounds grain2 -0.190 0.053 -0.109 0.030 Pounds hay2 -0.032 0.013 -0.039 0.016 pounds silagez -0.002 0.002 -0.016 0.017 Tot-ave no. of cows 553.856 360.409 0.015 0.009 1Regression coefficients and their standard errors for lbs. M.E. milkz, lbs. hayz, and lbs. silage2 must be divided by 10,000. 72 Appendix Table 2 Linear and quadratic, regression coefficients and beta weights of independent variables affecting herd average milk production in established herds when M.E. milk was included and with feed expressed as TDNl Regression Std. Errors Beta Std. Errors Coefficients of Wts. of Betas Coefficients M.E. milk 0.741 0.055 0.782 0.058 Ave. age 101.534 12.365 0.436 0.053 Ave. no. cows -3.446 0.966 -0.050 0.014 % cows in milk 447.746 91.226 0.782 0.159 Pounds TDN 0.227 0.053 0.184 0.043 Tot-ave no. of cows -l9.558 206.457 -0.001 0.010 M.E. milk2 0.031 0.019 0.092 0.058 Ave. age2 -0.808 0.112 -0.382 0.053 Ave. no. cows2 0.004 0.005 0.011 0.014 % cows in milk2 -2.222 0.533 —0.664 0.159 pounds TDN2 -0.116 0.033 -0.150 0.043 Tot-avg no. of cows 593.652 362.315 0.016 0.010 1Regression coefficients and their standard errors for lbs. M.E. milk2 and lbs. TDN2 must be divided by 10,000. 73 Appendix Table 3 Linear and quadratic, regression coefficients and beta weights of independent variables affecting herd average milk production in established herds when M.E. milk was not included1 m M Std. Errors Regression of Beta Std. Errors Coefficients Coefficients Wts. of Betas Ave. age 70.673 26.521 0.303 0.114 Ave. no. cows —20.979 2.165 -0.306 0.032 % cows in milk 564.581 195.394 0.986 0.341 Pounds grain 1.644 0.103 1.012 0.064 Pounds hay 0.142 0.034 0.147 0.035 pounds silage 0.050 0.013 0.147 0.038 Tot—ave no. of cows 310.482 443.641 0.014 0.021 Ave. age2 -0.685 0.241 -0.324 0.114 Ave. no. cows2 0.073 0.012 0.188 0.030 8 cows in milk2 -2.203 1.141 —0.659 0.341 Pounds grain2 -1.012 0.110 -0.582 0.063 pounds hay2 -0.068 0.027 -0.084 0.034 pounds silage2 -0.011 0.005 -0.073 0.037 Tot—avg no. of cows -362.301 778.373 -0.010 0.020 lReggession coefficiegts and their standard errors for lbs. hay and lbs. silage must be divided by 10,000. 74 Appendix Table 4 Linear and quadratic, regression coefficients and beta weights of independent variables affecting herd average milk production in established herds when M.E. milk was not included and feeds expressed as TDN Std. Errors Regression of Beta Std. Errors Coefficients Coefficients Wts. of Betas Ave. age 72.034 28.192 0.309 0.121 Ave. no. cows -23.980 2.180 -0.350 0.032 % cows in milk 845.629 207.231 1.476 0.362 pounds TDN 1.578 0.118 1.285 0.096 Tot—ave. no. of cows 537.723 472.057 0.025 0.022 Ave. age2 -0.753 0.256 -0.356 0.121 Ave. no. cows2 0.070 0.012 0.178 0.031 8 cows in milk2 -3.764 1.211 -1.125 0.362 pounds TDN2 -0.725 0.074 -0.943 0.096 Tot-avg. no. of cows —230.874 828.433 —0.006 0.022 1Regression coefficient and the standard error for lbs. TDNZ must be divided by 10,000. 75 Appendix Table 5 Linear and quadratic, regression coefficients and beta weights of independent variables affecting herd average milk production in new herds when M.E. milk was included1 w Std. Errors Regression of Beta Std. Errors Coefficients Coefficients Wts. of Betas M.E. milk 0.456 0.096 0.495 0.104 Ave. age 66.000 18.067 0.293 0.080 Ave. no. cows -4.353 1.144 -0.153 0.040 % cows in milk 361.714 159.114 0.648 0.285 Pounds grain 0.431 0.088 0.263 0.054 Pounds hay 0.075 0.040 0.075 0.041 Pounds silage -0.033 0.012 -0.099 0.035 Tot-ave. no. of cows 71.777 1056.917 0.003 0.047 M.E. milkz 0.109 0.034 0.329 0.103 Ave. age2 -o.520 0.165 -0.253 0.080 Ave. no. cows2 0.004 0.001 0.140 0.039 % cows in mi1k2 -l.680 0.942 -0.509 0.285 pounds grain2 -0.298 0.086 -0.184 0.053 pounds hay2 -0.073 0.037 —0.080 0.040 pounds silage2 0.013 0.005 0.086 0.034 Tot-avg. no. of cows 645.137 1975.715 0.015 0.047 1Regression coefficients and their standard errors for M.E. milkz, lbs. hayz, and lbs. silage2 must be divided by 10,000. 76 Appendix Table 6 Linear and quadratic, regression coefficients and beta weights of independent variables affecting herd average milk production in new herds when M.E. milk was included and the feeds expressed as TDN Std. Errors Regression of Beta Std. Errors Coefficients Coefficients Wts. of Betas M.E. milk 0.463 0.098 0.503 0.106 Ave. age 63.585 18.387 0.283 0.082 Ave. no. cows -5.568 1.060 -0.l96 0.037 8 cows in milk 406.603 161.780 0.729 0.290 pounds TDN 0.210 0.116 0.168 0.093 Tot-ave. no. of cows 46.113 1072.509 0.002 0.047 M.E. milk2 0.113 0.035 0.342 0.105 Ave. age2 -0.507 0.168 -0.247 0.082 Ave. no. cows2 0.005 0.001 0.179 0.037 8 cows in milk2 -1.956 0.957 -0.593 0.290 pounds TDN2 -0.098 0.071 -1.126 0.091 Tot-ave. no. of cows2 873.427 2002.778 0.021 0.047 lRegression coefficients and their standard errors for M.E. milk2 and lbs. TDN2 must be divided by 10,000. 77 Appendix Table 7 Linear and quadratic, regression coefficients and beta weights of independent variables affecting herd average milk production in new herds when M.E. milk was not in- cluded1 " fl _ L Std. Errors Regression of Beta Std. Errors Coefficients Coefficients Wts. of Betas Ave. age 77.714 36.610 0.345 0.163 Ave. no. cows -12.189 2.303 —0.429 0.081 8 cows in milk 281.484 321.601 0.505 0.577 Pounds grain 1.664 0.173 1.018 0.106 pounds hay 0.159 0.082 0.161 0.083 Pounds silage 0.038 0.024 0.113 0.071 Tot-ave. no. of cows 6708.322 2135.212 0.297 0.094 Ave. age2 -0.632 0.335 -0.307 0.163 Ave. no. cows2 0.011 0.002 0.378 0.079 8 cows in mi1k2 -0.545 1.905 -0.165 0.577 Pounds grain2 -0.979 0.171 -0.604 0.105 pounds hay2 -0.057 1.074 -0.062 0.081 pounds silage2 -0.005 0.010 -0.035 0.069 Tot-ave. no. of cows2 -11885.026 3990.619 -0.280 1.094 lRegression coefficients and their standard errors for lbs. hay2 and lbs. si1age2 must be divided by 10,000. 78 Appendix Table 8 Linear and quadratic, regression coefficients and beta weights of independent variables affecting herd average milk production in new herds when M.E. milk was not in- cluded and feeds expressed as TDN _ w _7 Std. Errors Regression of Beta Std. Errors Coefficients Coefficients Wts. of Betas Ave. age 77.002 38.510 0.342 0.171 Ave. no. cows -17.023 2.180 -0.598 0.077 % cows in milk 376.884 338.072 0.676 0.606 pounds TDN 1.526 0.236 1.217 0.188 Tot-ave. no. of cows 7078.226 2239.450 0.313 0.099 Ave. age2 -0.665 0.352 -0.324 0.171 Ave. no. cows2 0.015 1.002 0.531 0.076 8 cows in mi1k2 -l.064 2.001 -0.322 0.606 pounds TDN2 -0.658 0.145 -0.847 0.187 Tot-ave. no. of cows2 -11751.270 4183.313 -0.277 0.099 lRegression coefficient and standard error of TDN2 must be divided by 10,000. «III III I'll 'I'l II III 1 3