' ‘ DORIANSIRES . ; heslsforthenegreeofms j : 4- ' : ‘LMQ?HERVAS’ZORDONEZI: Ill'iifilflm'il WI]ll‘i'l‘l'lml'l‘lsl'li'l‘l'liil'lifulflifilsll 3 1293 o1§_68 61_28 . ABSTRACT RANKING 0F ECUADORIAN SIRES BY BEST LINEAR UNBIASED PREDICTION By. Thelmo Herwas Ordonez First and second lactation records from 785 Ecuadorian Holsteins distributed in four .altitude areas, three "year" groups and six "season" groups were analyzed with the objective of ranking sires. BLUP approach was used to rank sires from three models and computer'programs from Genstat were used to assist the analysis. Results show that sires did not rank the same in first and second lactation records. R2 statistics were obtained for each model; the highest value, 0.50, was obtained with a model (iii) which includes herd, year group, season group, and sire as sources of variation in first lactation milk yield. To compare ranking .of sires from ‘different models and lactations, Spearman's correlation was made. The highest correlation value, 0.98, was yielded between model (ii) which includes area, and model (iii) which includes herd; and between milk and fat yield. Although smallest 6a? was obtained in second lactation records, because a sire must be progeny tested early in life to obtain faster genetic improvement, ranking of sires using first lactation records in a model which includes herd, year, season, and sire is recommended in Ecuador. RANKING OF ECUADORIAN SIRES BY BEST LINEAR UNBIASED PREDICTION BY Thelmo Hervas Ordonez A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Dairy Science 1982 PREAMBLE . This work was initiated under the guidance of Dr. Ivan Mao, in the genetics. section of the Department of Dairy Science. The 58,455 records of milk production used in this study were provided by Dr. Charles Wilcox, university of Florida. The objective of the study was to learn to use dairy records to make certain management decisions and to gain experience using the computer. During the last half of my research Dr. Mao left for a sabbatical. After six unnths under the supervision of Dr. T. Ferris, my program was transferred to Dr. R.M. Cook. This thesis was completed under his guidance. Dr. William Magee and Dr. K.A. Wilson, in the Institute of International Agriculture also served as advisors, for which I am very grateful. ii To my father for his thought. To my mother for her faith. iii ACKNOWLEDGEMENTS I wish to express my gratitude: To Dr. Robert Cook for his assistance as Chairman of my Guidance Committee during the last period of my study program. To Dr. Ivan Mao, my advisor during the initiation of my graduate work, for his friendship, support and encouragement. To Dr. William Magee and Dr. Kim'Wilson, members of my Graduate Committee. To the people of Ecuador and Switzerland through the "National Institute of Agriculture and Livestock Research" (I.N.I.A.P.), and the "Swiss Technical Cooperation" Ihl the persons of Dr. Enrique Ampuero and Dr. Tony Rihs who made possible my studies here at Michigan State University. To Dr. Charles Wilcox from the University of Florida for providing the data used in this study. To Dr. Theodore Ferris for his help during the process of reading and editing the data. To my friends here at M.S.U. - Telmo Oleas, John Walter, Manuel Villarreal and Edgardo Cardozo - for their friendship. FinalLy to whom my heart belongs, my family, for their support and understanding. by TABLE OF CONTENTS Page Dedication . ........... ... ........... . .................... ... ............ iii Acknowledgements ........................................................ iv Table of Contents .... ...................... . ........................ ....v List of Tables ......................................... .........vii Chapter I Introduction ......................... . .................... ....1 Chapter II Literature Review ...... ..... . ..... . ............. . ........ .....3 11.1 Sire evaluation ..... ....... .............................3 .Background and historical development of sire evaluation.3 11.1.1 Factors considered in sire evaluation ............3 11.1.1.1 Environmental variations ................3 11.1.1.2 Environmental correlations ..............5 11.1.1.3 Seasonal effects ........................6 11.1.1. 4 Genetic trends ..........................6 II. 1.1.5 Age correction factors ..................8 11.1.1.6 Pedigree ................................9 11.1.1.7 Repeatability ...........................1l 11.1.1.8-Sire groups .............................11 11.1.2 Methods of sire evaluation .......................12 II. 1.2.1 Daughter average ........................13 11.1.2.2 Daugher-Dam comparison ..................13 11.1.2.3 Herdmate comparison .....................13 II. 1.2. 4 USDA-DHIA modified contemporary compari- son (MCC) ..... ........ . .......... . .16 11.1.2.5 Cumulative difference ........ .. ......... l7 11.1.2.6 Sire evaluation by linear models ..... ...17 11.1.3 First lactation records sire evaluation .. ..... ...22 11.1.4 Later lactation records sire evaluation ...... ....23 Chapter 111 Materials and Methods .. ....................................... 27 111.1 Materials ..... . ......... . .......................... ....27 111.1.1 Sources of data ................................ 27 III.1.1.1 Origin ........................... ....27 111.1.2 Screening of data .............................. 27 111.1.3 Sire groups .................................... 28 111.2 Methods ................. ........... . ....... ............ 28 111.2.1 Models used .................. _ .................. 47 Chapter IV Results and Discussion .........................................50 1V.1 Records management .... ..... .......... ........ ............50 1V.l.1 Age ...................... . .................. ......50 1V.1. 2 Date of birth ..................................... 50 1V.l.3 Freshening date . ......... ...... ........... . ....... 51 1V.l.4 Lactation length ... ........................ .......51 1V.l.5 Conditions affecting the record ................... 51 1V.l.6 Area . ............................................. 51 1V.l.7 Type of animal and body conformation .............. 52 1V.l.8 Lactation number ..................................52 1V.l.9 Milk production . ...... . ........................... 53 1V.2.1 Examination of factor effects .......................... 54 1V.2.1.1 Area ............................................ 54 1V. 2.1.2 Herd ...................... .......... ............S4 1V.2. 1. 3 Year group .................... . ............... ..54 1V. 2. 1.4 Season group .................................... 54 , 1V.2.1. S Interactions .................................... 50 1V.2. 1. 6 Sire ... ......................................... 60 1V.3 Sires Rank ...... .. .............. . ................ ......60 Chapter V Summary and Conclusions ........................................ 33 Bibliography ..................................................... 85 Page vi Table 111.1 IIIOZ 111.3 111.4 111.5 111.6 111.7 111.8 111.8.1 111.9 111.10 111.11 111.12 111.13 111.14 LIST OF TABLES Page Cow's age in years. Number of observations, percentages relative and cumlative .......0000...OOOOOIOOOOOOOOOOOOOOOO0.00... 29 Cow's birth date by year. Number of observations, percentages relative and cumulative ......OOIOCCOOO ..... ......OOOCOOIOOOOOOOOOO 3o Cow's birth date by month. Number of observations, percentages relative and cumlative 0.0.0.0000...0............OOOOOOOOOOOOOO... 31 Cow's freshening date by year. Number of observations, percentages relative and cumlative ......OOOOOO....OQOOOOOOOO.....OOOOOOOOOOOO 32 Cow's freshening date by month. Number of observations, percentages relative and cumlative .00.......0.0.0.000...OOOOOOOOOOOOOOOOOOOOOO 3 Cow's lactation length in months. Number of observations, percentages relative and cumlative 00.0.0.........OOOOIOOOOOOOOOO0.0.0....0...... Cow's condition affecting the record. Number of observations, per- centages relative and cumulative ....................................95 Area description 00......I.0....O....IOOOOOOOOOOOOOOOOIOOOOOO00......36 Cow's distribution by area. Number of observations, percentages rela- tive and cumlative .0.........0.00.0000...0.0.0.0000.........OOOOOOO Cow’s type. Number of observations, percentages relative and cumula- tive ......OOOOOOOIOO......OOOOOOOOOOOOOOOOOO......OOOOOOOOO0.00.00.0037 Cow's body conformation classification. Number of observations, per- centages relative and cumulative ..................... ...... .........37 Cow's lactation number. Number of observations, percentages relative and cumUIative ......OOOIOOO000......I.......OOOOIOOOOOOOOOOOOO.....0038 Cows in area. Frequencies absolute, relative and cumulative ......... 39 Freshening year group at first lactation. Frequencies absolute, relative and cumulative ........ .... ........ . ......................... 40 Freshening month group at first lactation. Frequencies absolute, relative and cumulative .............................................. 40 vii Table 111.15 111.16 111.17 111.18 1V.1 1V.2 IV03 1V.4 1V.5 1V.6 1V.7 1V.8 1V.9 1V.10 1V.11 1V.12 1V.13 1V.14 1V.15 Freshening year group at second lactation. relative and cumulative ..... Freshening month group at second lactation. Page Frequencies absolute, 41 0000000000000000000000000000000000000 Frequencies absolute, relative and cumlative 00000007000000000000000000000000.0000000000 41 Farm Sire Milk Milk cords . ......... .... 2 distribution. Frequencies absolute, relative and cumulative..142 distribution. Frequencies absolute, relative and cumulative. 43 yield, fat percentage and fat yield. Actual records ........ 55 yield, fat percentage and fat yield. Mature equivalent re- ............... ........ . ............. ......... 56 R statistics for the different models and "Y" variables in first and second Summary of Summary of Summary of Summary of Summary of 57 lactations 000000000000 0000000 analysis of variance in first lactation. Area model (i).57 analysis of variance in first lactation. Area model(ii) 53 analysis of variance in second lactation.Area model(iii) 58 analysis of variance in first lactation.Herd model (ii£)59 analysis of variance in second lactation.Herd model(ii£)59 Range of BLUP estimates and standard error in models (ii) and (iii) for milk yield, fat Percentage, and fat yield, in both first and second lactations 000000000000000000000000...000000000000000000000062 Spearman's rank correlation for models (ii) and (iii), all traits and lactations ... Rank (11) Rank area Rank (11) Rank (11) Rank area .. ..... 63 of sires, lactation one, according to milk yield. Model area andherd (iii) 00000000000000000000000000000000000 00000 000000065 of sires, lactation one, according to fat percentage. Model ~(ii) and herd (iii) ........ ..... .... ..... .......... ........ .. 68 of sires, lactation one, according to fat yield. Mbdel area and herd (iii) .. ..... 71 of sires, lactation two, according to milk yield. Model area and herd (iii) 0 0 0 0 0 0 0 0 0 0 0 0 0 00000 0 0 0 0 0 0 0 0 O 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 74 of sires, lactation two, according to fat percentage. Model (ii) and herd (iii) ............................. .. ........... 77 Table Page 1V.16 Rank of sires, lactation two, according to fat yield. Model area (11) and herd (111) ............................................. 80 CHAPTER 1 INTRODUCTION Dairying .is one of the top enterprises. in Ecuador's agricultural industry. Dairy farms in Ecuador are found mainly in the Andean region from 2,250 to 3,400 meters above sea level. Plenty of sunshine, rainfall distributed throughout the year, moderate temperatures ranging from 10 to 17°C, and vigorous grass growth in this region create ideal conditions. for milk Production. The predominant breed of dairy cattle is Holstein-Freisian. Registered animals have been imported since the beginning of this century and native cattle have been upgraded for several generations through artificial insemination, principally using U.S.A. proved sires. Dairy scientists have been working to increase milk product- ion through genetic progress and enhanced environmental conditions. Research on genetic and environmental factors affecting milk production, genetic and environmental trends, correction factors, and production parameters have been done. New methods, techniques and improved management systems have been tested and adopted when convenient. The selection of genetically superior parents has a great influence upon the improvement of succeeding generations. This study evaluated sires in Ecuador using the BLUP approach method of sire evaluation. The information obtained should aid sire selection and genetic improvement of the dairy cattle population in this area of the world. CHAPTER II LITERATURE REVIEW 11.1 Sire Evaluation Background and historical development of sire evaluation The rate and success of genetic progress is dependent upon the selection of sires whose daughters will produce 'milk at a more profitable level than their dams. About 902 of the pressure to improve potential in the North American population is on the sires (McDaniel, 1974). Because of the essential role that sires 'play' in. genetic improvement of dairy cattle, researchers have sought more effective methods of sire evaluation. Estimates of the genetic worth of bulls have been produced through the consideration of factors chosen to make these estimates as representative as possible of the genetic transmitting ability of the sires (Norman, 1974). With emphasis on certain traits, such as udlk production, many factors must be considered in order to optimize our results. 11.1.1 Factors Considered in Sire Evaluation II.l.l.l Environmental variations Johanson (1960) working with progeny testing methods, found that for. an accurate evaluation of sires it is necessary to eliminate the effects of systematic environmental factors that are affecting milk production. The method used in this process is to compare the production of the bull's daughters. 3 Keown (1974) writes that environmental variation has existed and probably still exists. He found that highest yields were obtained in January-February and were lowest for July-August calvings. McDowell et a1. (1976) working with Mexican data found that the effects of. the year, the season, and the age of the cows. were important influences on milk. yield. The variation in, milk. yield accounted for by four climatic variables plus feed and body weight ranged from 11 to 62%, the larger variation being for cows calving in July and August and the smaller for those calving in spring and fall. Climatic conditions appeared to have the greatest influence in the first 60 days of lactation. Roman (1970) using Ecuadorian data, pointed out that seasonal effects were not important sources of variation affecting milk and fat yield as is sometimes the case in temperate zones; the variability due to season of freshening was less than 1% of total variance in registered or grade cows. Season classification was January to March, April to June, July to September and October to December. A general conclusion from experiments on the effect of climatic stress is that the best yields and efficiency of performance could be obtained under stable environmental conditions with temperatures in the comfort between regions in the Predicted Differences (P.D.) of bulls. The mean of herdmate sires' P.D. should be .approximately equivalent to the average genetic transmitting ability value of the bulls used.:u1 the area previous to the period studied. He indicates that when using AI in the same breed for several years, the variation in production within a farm is from environmental conditions. The changes in production that are reflected in the herd averages are caused by environmental factors, management, sanitation practices, and nutrition. Roman (1970) analyzing Ecuadorian data, found that farms within area accounted for 26.32 and 37.5% of variability in milk yield, and 5.90% and 15.65% for fat yield, in registered and grade cows respectively. 11.1.1.2 Environmental correlations Several authors (Bereskin and Freeman, 1965; MCDaniel, 1974; McDaniel and Plowman, 1961; Thompson and Freeman, 1970; and Arora and Freeman, 1971) have reported small environmental correlations in different populations ranking from near 0 to 0.14. These authors agree that there are a number of possible causes of environmental correlations. Among these are: a. Failure to remove all herd effects, herd. year effects, year-season effects, or herd-year-season interaction effects. b. Paternal half-sisters may be managed and fed more alike than other cows in the same herd-year-season. Large numbers of progeny would tend to eliminate environmental biases if a bull's progeny were compared to a random sample of cows. There are high correlations between herdmates' sires :h1 repeated samples. Environmental biases would probably be small if there were no correlations between a bull's breeding value and that of the sires of his progeny's herdmates. Evidence suggests a strongly positive association. 1V.l.l.3 Seasonal effects Even though many sources of variation have been considered in sire evaluation including herds, regions, production level, .lactation .length, days open, genotype by environmental interactions and others, one of the more important factors which ought to be considered is the seasonal effect on milk yield. In Michigan, Wunder and McGilliard (1971) found differences in milk yield between seasons of calving and age of animals. IV.1.1.4 Genetic trends Changes with time are due to both genetic and environmental factors and trends in a population or in individual herds are difficult to disentangle. warwick (1979) found. genetic trends are of interest when comparing sires used in different periods of time. van Vleck (1961) noted that genetic trends and overlapping. generations create difficulties in comparisons between younger and older sires. Differential use of sires by dairymen has caused the progeny of some bulls to be compared with herdmates that are better genetically than herdmates of other bulls. There are positive genetic trends in Artificial Insemination (AI) sires indicated by both udlk and fat yields in dairy cattle (Everett, 1976). The annual changes in management and genetics showed that in recent years there were more improved genetic evaluation procedures. With continued economic restrictions on feed input, breeding is expected to continue as an important factor for improved. yields (Powell, 1979). Lush and Shrode (1950) studying selection or culling, genetic or environmental time trends and repeatability, pointed out that positive genetic trends could be confused with age, and to determine the size of these biases, additional information regarding the amount and kind of selection practices is needed. Positive genetic trends make herdmate} comparisons for young bulls, relative to older bulls, lower than they really are (MCDaniel, 1973); Genetic trends are generally obtained by comparing overall production. with environmental estimations of milk or fat yield (Burnside, 1967). Mao (1971) using age-month adjusted records, pointed out that time trends may be the result of genetic homogeneity and/or environmental heterogeneity. Genetic trends have little importance in evaluation of contemporary sires. If sires used in widely different time periods are to be compared, trends in genetic merit must be considered. Least squares procedures and family relationships have been used used to estimate genetic trends (Van Vleck and Henderson, 1961; Smith, 1962). Rodriguez (1974) has detected in an Ecuadorian Holstein population a positive genetic trend in yields and negative environmental and phenotypic trends in yields. In Mexico, positive genetic trends for sires from the USA, Mexico and Canada were found by McDowell et a1. (1976). Adkinson (1972), using Ecuadorian data, dated from 1964 to 1968, found that genetic trends for yields were positive and curvilinear, appearing to plateau around 1960. Trends were linear and positive for fat percent and accounted to 0.02410.0047. per year; he found that environmental trends were negative and curvilinear for milk and fat yields, negative but linear for fat percent and negative environmental trends which might appear unusual but have been reported previously. 11.1.1.5 Age correction factors Age [of calving is one of the main factors affecting milk. and fat yield in dairy cattle. Comparisons between cows frequently include animals of different ages, and since cows generally produce less when young and during old age, it is necessary to standardize records with age correction factors developed for that purpose (Lush and Shrode, 1950; Mao, 1974). The age distribution of both daughters and herdmates may differ among sires. It is generally assumed that application of age correction factors will remove some biases; however, other effects are confounded with age such as herd—year, cow selection, breed and geographical region; ltherefbre, age adjustments ought to be made either jointly with the adjustment for these combined effects or free from.them.(Mao, 1974). There is some controversy regarding benefits of using cows of all ages in sire evaluation. Sire sampling is by necessity concerned ‘with first lactation daughters (Powell, 1972). Changes in lactation yield with season of calving and age were found by wunder and McGilliard (1971). Mao (1973), working with first and second lactations, found differences in lactation yield at calving ages between 31-37 months. The same author (1974) recommends age month adjustment factors for accurate comparisons of cow's productivity. A. method known as. Gross Comparison was developed by' Gowen (1920,1924). 'It establishes a comparison of milk production averages of different age groups. Sanders (1928) in an attempt to diminish the biases of this Gross Comparison Method, weighed each consecutive pair of- records between the same group of cows, thus creating the Paired Comparison Method. Later, Beardsley (1952) used a maximum likelihood (ML) procedure to create an age correction factor considering herd and age effects. Lush and Shrode (1950), Searle and Henderson (1959), and Mahadevan (1951) discussed weaknesses and biases of these methods. Miller and Henderson (1968), working with age correction factors ' pointed out that ML estimates may or may not be biased depending on the aptness, appropriateness and completeness of the model. Roman (1970), found in Ecuadorian Holsteins that within herd variability in ‘milk yield and fat yield was due to age variations. He reported that seasonal effects were not an important source of variation in milk and fat yields. Age correction factors developed by Rodriguez (1974) in an Ecuadorian Balatedn-FrieSian population, were found to be very similar to those developed in the temperate zones of the United States of America. 11.1.1.6 Pedigree A sire's proof is more reliable when pedigree is included. It is a record of the animals from which a given individual is descended. It includes identification of ancestors, collateral relatives and information on their performance (n: progeny ‘records (warwick, 1979). 10 Normally, a pedigree index is done using a multiple regression approach. In general, the contribution of information on a relative toward estimated pedigree index for the bull in question is a function of three factors. a. Relationship of the relative to the individual; b. Relationship of the relative to the other relatives used in the evaluation; and c. Accuracy of the ‘breeding value estimate on the relative (Butcher, 1967). A pedigree index is useful as a relative estimate of a bull's breeding value. Combining information on sire, maternal grand sire, and dam's early lactation by current pedigree indexing procedures is effective for screening prospects .for a young sire sampling program (Butcher, 1973). Van Vleck and Carter (1972), working with Holstein bulls, concluded that pedigreeindexes were an effective method of selecting young sires and Casell et a1. (1976) added to the evidence of benefits from careful pedigree.se1ection for production traits in bulls used under limited and multi-herd conditions. Butcher (1976) pointed out that high pedigree index bulls have a much higher probability of achieving a high P.D. proof. The maximum rate of genetic improvement in yield traits in a closed dairy cattle population is obtained when an appropriate proportion of the mating is to young bulls selected on pedigree (Butcher 1976; Specht, 1960). 11 11.1.1.7 Repeatability The regression of future performance on past performance is called repeatability. It is really a confidence factor that depends on: 'a. ’How many daughters are available; b. How many herds they are in; and, c. How they are distributed among herds. Repeatability is a factor which ranges from 0 to 992 depending upon the amount of daughter information and distribution of daughters over herds. The repeatability is higher when more daughters .are randomly distributed over more herds. The higher the percent repeatability, the more accurate the P.D. value and the narrower the confidence range in the P.D. milk (MtDaniel, 1974; Mao, 1980). 11.1.1.8 Sire groups Ideally, groups should be defined alike so there is maximum genetic similarity within sizable groups and in such a way that the means of groups represent genetic differences. Possible bases for grouping are year of birth, year of first daughter freshening, year within stud or' pedigree estimate (Powell and. Freeman, .1974; McDowell et a1., 1976). Grouping by geographical area is beneficial because the genetic value of a bull could be affected by the geographical area in which their daughters are milked; herds, regions or areas with higher selection differentials on bulls have the most under—rated. bulls (McDaniel, 1974; Tomaszewsky, 1973; Henderson, 1974). 12 When we. group sires, genetic trends are accounted for if the estimates of merit for sires of herdmates are unbiased and properly applied. Working with a mixed model without groups accounts for the merit of sires of herdmates, but the estimates of merit for all sires are biased by regressiOn to an inappropriate mean. Grouping is an arbitrary process. The purpose of grouping sires in a model is to recognize that sires are not all random samples from a static population. Grouping is an attempt to identify and take into account‘the non-random source of the sire sampling. Grouping sires by pedigree could be more effective in removing the environmental variation than grouping by stud-year as is done by the Northeast Artificial Insemination Sire COmparison Program (NEAISC), if there are differences in quality of bulls purchased each year by individual studs (Norman, 1974). I Pedigree grouping would encourage the. sampling of sires with outstanding pedigrees because the estimate of an individual sire is influenced to a large extent by the group mean, particularly when evaluated on a limited number of daughters (Norman, 1974). 11.1.2 Methods of Sire Evaluation The goal of breeders is to obtain genetic estimates of bulls which account for independent non-genetic factors. Those estimates are as representative as possible of the genetic transmitting ability of sires (Normal, 1974). Sire evaluation methods have been developed from genetic theory and usually include simplifying assumptions. While the approximate validity of the assumptions used is subject to experimental l3 verification, the inability to measure true genetic value complicates empirical comparisons among methods (Jamison, 1977). Over time, many procedures have been developed. In general, any procedure generates estimation of some sort and only the differences among breeding values are meaningful (van Vleck, 1976). 11.1.2.1 Daughter Average The Daughter Average is the simplest method of sire evaluation (the average production of the daughters of a bull computed to the average production of daughters of other bulls). 1t once had as its prerequisite the following: The compared bulls must come from the same population and there must be genetic equality among mates of the bulls (Van Vleck, 1976). 11.1.2.2 Daughter-Dam Comparison The Daughter-Dam Comparison method (trying to measure the effect of a bull by comparing dam with daughter), was suppose to account for differences in mates but actually may have been more useful in adjusting for herd level. One of the biases of this method is that it puts too much weight on dam average so one of the weakest links is the change in herd-year-season effects from dam to daughter (Van Vleck, 1976). 11.1.2.3 Herdmate Comparison (HMC) The Herdmate Comparison method, developed by C.R. Henderson and A. Robertson in 1950 is the well known fixed model which was computed using Least Squares (LS). 14 Where: "Y" is an observation vector (records); "X" is known matrix; " " is an unknown fixed vector; and "E" is a more observable random vector with a mean, a veCtor of zeros and a1 variance-covariance matrix (Henderson, 1974). It does contain.21'herd-year-season effect. This method uses a difference between a sample of the transmitting ability of a bull and the average transmitting ability of sires of the daughter's herdmates plus the difference between her dam's transmitting ability and the average transmitting ability of dams of her herdmate. Several assumptions underlie this method. For example, the sires of herdmates of daughters of bulls are equal in average transmitting ability and in the genetic base of the group to which the sire belongs. Other assumptions cannot be supported. These include. 1. Bulls'Fdaughters do ‘not receive preferential treatment relative to herdmates; 2. Each bu11:is mated to a representative sample of the population for the traits evaluated in the progeny; 3. A11 sires with tested daughters from a breed are a random sample from a single population; and, 4. The distribution of sires across the herds is random. The Predicted Difference (P.D.) will drop each year, on the arverage, by an amount equal to one-half of the yearly genetic improvement in the dairy cow population. This is the reason why the IP.D. for older bulls appears somewhat more favorable compared to the P.I). for younger bulls than it actually is (Van Vleck, 1976; McDaniel, 15 1974). The P.D. is defined as the expected deviation in milk or fat of a bull's progeny from their herdmates in a breed-average herd. It compares a bull's daughter to the progeny of other bulls that calved in the same herd at the same time, taking into account season and age at calving and length of lactation. Besides having higher milk yields, it has also been shown that the daughters of high P.D. sires: 1. Have more income over feed cost per lactation; 2. Have greater feed efficiency; 3. Voluntarily consume more forage when it is offered free choice; and, 4. Have longer productive lives than daughters or lxmv P.D. sires. The USDA sire summary, utilizing herdmate comparisons,. is accurate enough to be considered an effective tool for genetic improvement and it is widely used in dairy cattle (McDaniel, 1974). Norman et a1. (1972) have shown that the deviation of a progeny from herdmates depends on .the genetic ability of the sires of herdmates. As the genetic value of the herdmates increased, so did daughter average yield, but daughter deviation from herdmate average decreased. There are several reasons why the herdmate comparison is inadequate. Sires of herdmates from progeny of specific sires will differ in genetic merit because sire proofs tend to decline in successive generations. This can produce wide discrepancies in merit of sires of herdmates among sires being tested (Powell, 1974). The same author points out that the classic ‘method of sire l6 evaluation by herdmate comparison does not account for the merit of sires of herdmates. P.D. is affected by daughters' distribution across herds because of the sires' repeatability is used to regress the daughter-herdmate deviation toward the population mean (Norman, 1974) . The P.D. method places great reliance on accurate age factors since young cows are compared with herdmates of all ages and old cows are compared with herdmates of all ages (Everett and Henderson, 1972). 11.1.2.4 USDA-DHIA modified contemporary comparisOn (MCC) This method was implemented in 1974 to replace the HMC in order to overcome the invalidity of some assumptions (USDA Report, 1976). For the MCC, the contemporary average encompasses ‘records initiated during a five month interval. It includes animals which calve two months before the individual daughters calve, the month the daughters calved and two months after the daughters calved. In order that the contemporary comparison be effective the sire's daughters and contemporaries must be provided the same opportunity (Warwick, 1979). The MCC utilizes two contemporary groupings: 1.First lactations (MCl); and, 2.Second and later lactations (MCL). First lactation daughters are compared to M01 and. second and later lactations to MCL. The results of the MCL are expressed as a Predicted Difference (P.D.) which is an estimate of the expected average of many future daughters of the sire. 17 The major improvement in the MCC is the adoption of a genetic base from which all P.D.s are expressed to minimize the impact of genetic trends :hi the comparison of bulls over time when the MCC was implemented _in 1974. The genetic ‘base ‘was related. to the average calving dateof the records used in the fall 1974 USDA Sire Summaries. This base. can be changed in accord with the average genetic change that has taken place in the breed (USDA Report, 1976). 11.1.2.5 Cumulative Difference The Cumulative Difference (CD) is another important method developed by Bar-Anan and Sacks (1974). It offers a prediction of breeding values of sires when data are available only on progeny. The usual approach is essentially a two-way classification model in which unbiased estimates of progeny means are obtained and then' breeding values of sires are found by weighting these estimates by their heritabilities. The evaluation becomes available over a period of time and the information of estimation is valid during that time (Thompson, 1976). Bar-Anan (1974) points out that although a daughter record appears in only one period of time, sire tests in different periods are not independent since there is a covariance between the records of a bull's daughter in different periods. 11.1.2.6 Sire evaluation by linear models Mixed linear model methods for sire evaluation have been developed by Henderson and they provide a powerful tool for 18 use under a wide variety of situations (warwick, 1979). The general form of the mixed model equation is: Y I x B + 23 + e Where: Y is a vector of observations; 8 I 'is a vector of unknown fixed effects; . s is a vector of unknown random effects; e is a random residual effect; x, z are design matrices; and, E(e) a 0, var (e) - r28, E(e) a 0, var (e) = r2. This method, originally called Maximum Likelihood (ML), has been used in breeding research for more than 25 years (Shaeffer, 1976). The Northeast Artificial Insemination Sire Comparison (NEAISC) is a mixed linear model method developed by Henderson at Cornell university. It uses records from.two year old freshening cows as a genetic base and has several advantages. On method The bulls are evaluated according to the levels of competition It eliminates problems of differential culling among sires The age correction factors problem is reduced It eliminates the biases caused by genetic trends It uses sire groups (Everett and Henderson, 1972; Everett and Quaas, 1979; Everett, 1974). November 15, 1966-, Dr. C. Henderson introduced another of sire evaluation based on statistical concepts which are referred.tx> as the BLUP techniques. It was developed for working 19 with selection indexes and linear model techniques to deal with a large set of data with unequal subclass numbers. This new method accounts for unknown and environmental trends, herd, season, age effects and differential culling of daughters. This method has no statistical limitations except computer storage and ability to handle large untrix operations for a specified model and all the related assumptions and restrictions (Mao, 1980; Shaeffer, 1975). A simple modification of the regular least squares leads to a BLUP solution. To the evaluation problems in situations of unequal numbers R.- I, G.1 is simply added. When the sires are unrelated and herds are fixed, this means addingorezlrzS to the diagonal of the sire equations; this ratio in heritability terms becomes (4 - h2)/h2, (Van Vleck, 1976). If R i 1 some modifications are required. The equations in this case are solved by the Gauss-Siedel iterative method. As with any. statistical procedure, the validity of its properties depends on the development of a correct model to describe the data. If the data is appropriate then the model as noted by Henderson (1975), Van Vleck (1976), Mao (1979, 1980), Vinson (1979), Casell (1979), and Shaeffer (1975) has the following characteristics. 1. Can be used in a wide variety of situations 2. The evaluation obtained is unbiased and has the smallest possible variance of prediction errors 3. Combines unequal numbers of daughters unevenly distributed in a population of herds in an optimal way ' 10. ll. 12. 13. 20 The correlation between predicted and true variance is minimum. 1f the data are true values and have a multi-variance normal distribution the probability of correcting and ranking pairs of true values is minimum. The maximization of genetic progress through truncation selection. Maximization of the probability of selecting the better of two bulls. Pairs of sires which do not have direct comparison may have indirect comparisons with common sires. Maximization of the correlations between predicted and predictor. Can use pedigree information and other factors, making this method a super machine that is very flexible and eliminates bias from genetic trend. Considers pure genetic uakeup between sires by comparing daughters of different sires directly or indirectly. Probability or confidence statements about genetic values can be made from the predicted true values and variance errors of prediction which come out directly in the Arithmetic of Predications. Sire solutions within group sum to zero. Some mathematical disadvantages of BLUP include the following. 1. Only linear functions of observations are used in common with nearly every other procedure. Although unbiasedness is a desirable characteristic, certain biased procedures may have a smaller squared prediction. error, but there is no way to tell if the biased procedure results in more or less prediction errors than BLUP; The correct ratio of variances of random effects (sires) is assumed known for predicting genetic values; it depends on heritability which is often known within reasonable limits. 21 4. If an equation is needed for every level or every factor in the model in animal breeding, problems could result in thousands of equations and unknown quantities to be estimated or predicted. 5. Records must be standardized or adjusted by age to avoid biases produced by unbalanced data. 6. Efficient absorption and sorting procedures are necessary in solving BLUP equations (Van Vleck, 1976; Mao, 1980; Ufford, 1977; Shaeffer, 1975). BLUP approach has been used in many ways. For example, it could be used in estimation of genetic trends (Miller, 1970; Lentz et al., 1969; Everett, 1972; Henderson, 1973) with categorical data (Conolly, 1981) and, of course in sire evaluation for production and type characteristics (Welter and Mao, 1981). It is not known at the present time what are the more appropriate Computing strategies and procedures (Henderson, 1974; Ufford, 1977). Evaluation by BLUP methodology require the solution of a large set of linear equations. It requires knowledge of LS methods and computing. strategies as ,the well-known absorption technique ‘(Ufford, 1977). Computing alogrithms for several models are available for. sire evaluation by BLUP' (Henderson, 1966). The procedures for solving equations are very flexible but could become complex according to the number of variables present in the model (Mao, 1980). It is a necessary mathematical procedure to reduce the number of equations to a manageable limit and still obtain the same solution as if all the equations 'were solved (Shaeffer, 1975; ‘Ufford, 1977). 22 To obtain solutions some restrictions must be imposed. Any two nonestimatable restrictions will result in. solutions *which. will estimate the same estimatable functions. Solutions to the random effects will always be the same, but solutions to the fixed effects will depend on the restrictions used. The expected values of the solutions. will, however, indicate what functions of the fixed effects can be estimated. The expected values of the solutions can be obtained by umltiplying the generalized inverse of the BLUP by the original least square equations and solutions can be obtained for multiplying the inverse times the totals on the right hand side (Van Vleck, 1976; Mao, 1980; Shaeffer, 1975). 11.1.3 First Lactation Records and Sire Evaluations First lactations of dairy cows are of special importance to sire selection as the most frequent lactation. They are 24 to 282 of all lactations (Powell, 1972). Powell (1972) suggest that the first lactations should be the most meaningful because previous production selection and cumulative disorders are not attributable to a cow's merit. Dairy cattle breeders have pondered. whether information from first lactation production is accurate enough as a predictor of performance in later lactations. This would allow sire evaluation to be based only on first lactations (Tomaszewsky, 1974)._ Butcher (1968) pointed out that there are several possible explanations for lower utility of later lactations of the cow: a. Preferential treatment relative to herdmates b. Lower heritability of later records c. Small genetic correlations between lactations 23 d. Effective repeatability of less than 0.5 Every attempt should be made to compare equivalent lactation records; otherwise serious problems may result. If a cow that has a first lactation record is compared with another cow that does not, a serious bias is introduced. This is due to a first lactation record being compared with a selected set of later records (Keown, 1976). 11.1.4 Later Lactation Records and Sire Evaluations Second selection tests are not only important for asserting sire lactation number interactions but also for increasing the accuracy of the estimated breeding values. By combining first and second lactations in progeny test proof, repeatability is increased and a yield increase per cow year in daughters of proven bulls is obtained (Bar-Anan, 1975). Casell and McDaniel (1980) working with first and all lactations in sire evaluations concluded that: 1. Proofs for milk based on later lactation records exceeded proofs based in first records. 2. Sire differences in differences between. later' and. first proofs were noted. 3. Correlations between first and later lactation proofs were low enough to produce some misranking of sires for later lactation progeny tests. 4. Increased culling on first lactation performance resulted in increased differences between first proof and later proofs. Comparing a sire's daughter to other cows in the herd, calving in the same period has become accepted in evaluating bulls. Some sire summaries 'have compared first lactation daughters to first lactation contemporaries. The present USDA-DHIA (United States 24 Department of Agriculture-Dairy Herd Improvement Association) sire summary compares progeny of all ages to herdmates of all ages (Norman, 1974). The belief that sires differ in the rate that their daughters mature seems to be a major reason why some people prefer the use of all lactation records. Differing rates of maturity may be accounting for a sire by age interaction. That/is, two sires may transmit the same merit for a first lactation period but yet differ in merit for a second lactation period (Ufford, 1977). Nicholson et a1. (1974) evaluating sire by age interactions using only first, only second, only third, only fourth lactation records having adjusted records for selection and using BLUP procedures, found that a few bulls did differ significantly in their evaluation from one lactation to the next. They concluded that the first lactation proofs were an accurate predictor of later lactation proofs. If the time arrives that sire by age interaction should be considered, each lactation could be treated as a separate trait correlated with other lactations being necessary to weigh each lactation according to its economical importance (Ufford, 1977). The residual error of an individual sire seems to be of special importance for evaluating breeding programs which use regression parameters for sire evaluation. This is because the animal breeder 'wants to know' what range in the indirectly estimated. daughter performance he may expect from using an individual bull (Bar-Anan, 1974). In a herdmate sire evaluation, Tomaszewsky et a1. (1975) 25 concluded that bulls ranked similar for first or second evalutions. The use ‘of later records in a sire summary is based upon the assumption that the same genes are influencing both first and later lactations. If this assumption is not justified, it may be that more emphasis should be placed in later lactations (Wickham, 1976). Research done by Shaeffer (1975) and Wickham (1977) support the statement that later lactations are influenced by a different set of genes. Therefore, there should be two sire summaries: one based on first lactation information and the other one based on later information records. Nbrman (1974) reports that the accuracy of sire summaries are affected by the number of cows to which each daughter is compared and the genetic correlations between yield in various lactations. The use of multiple records will result in a substantial increase in the time required for computing sire summaries. A thorough analysis of sire evaluation should consider the following points: 1. The reduction in sampling variance from adding daughter and additional herdmates in comparison. 2. The estimates from the literature of genetic correlations between first and later lactation records. 3. The value of additional accuracy compared to the cost of acquiring such information. When comparing selected later records with first lactation records, a problem arises: if all records are used, some adjustment must be made for cows with later lactation records that do not have first lactation information. If the records are adjusted properly, 26 the addition of these records in BLUP methods will help estimates of sires (Keown, 1976; Norman, 1976). The merit of incorporating all lactation records can be considered in terms of prediction error variance and bias. Since the £11}ch lactation records are- .a subject of all lactation records of prediction, error variances will be. smaller when all records are used. Unfortunately, biases are difficult to evaluate (Ufford, 1977). Use of multiple records appears to be more acceptable to the dairy industry. CHAPTER III MATERIALS AND METHODS 111.1 Materials 111.1.1 Source of Data 111.1.l.1 Origin The data set analyzed in this study was obtained from the official milk testing program of Ecuador through the university of Florida. The original data set contained 58,455 records from 1948 to 1967. Each record contains the following coded information: farm, breed, cow, sire, dam, birth date (month and year), freshening date (month and year), age in months, condition affecting records (CAR), lactation length, times milked per day, area, cow's classification (pure breed or grade), milk production in pounds, classification by type of body conformation and lactation number. The statistical package for social sciences (SPSS) was used to obtain. some statistics, frequency tables and milk yield, fat percentage and fat yield means; tables 111.1 to 111.11 summarize the results. 111.1.2 Screening of Data Fortran programs were written to read, clean, and edit the data. During this process records were deleted for the 27 28 following reasons: - Missing information on some variables: farm, cow, sire identification, date of birth, date of freshening, age, conditions affecting the record, incomplete lactation or lactation number. - No information of milk yield, fat percentage, and fat yield. - Sires having less than two daughters. - For parity, cows older than 48 months of age at first parturition. The final data set contained information from 101 sires, 785 cows with both first and second lactation records in 30 herds, four geographical areas, three year groups, six season. groups. Each record was standardized to a mature equivalent age using values developed with the original population and a maximum likelihood model which includes farm, cow, freshening date, linear, quadratic. and cubic length of record; and linear, quadratic, and cubic cow's age as source of variation (Rodriguez, 1974). 111.1.3 "Sire Groups The grouping criteria used for the different statistical models were: area, herd within area, five year group, and two month season. 111.2 Methods To rank the sires the BLUP approach was used and the general statistical program (Genstat) was used to obtain solutions. 29 TABLE 111.1 Age of cows in years, number of observations, absolute relative and cumulative frequency for the original data set Age Absolute Relative Cumulative frequency frequency frequency .(year) (PCT) (PCT) 1 t 12“ t 0.02 0.02 2 497 0.85 0.87 3 7965 13.62 14.49 4 8114 13.88 28.35 5 6902 11.80 40.16 6 5864 10.03 50.19 7 4606 7.87 58.07 8 . 3559 6.08 64.16 9 2623 4.48 68.65 10 1977 3.38 72.03 11 1287 2.20 74.23 12 778 1.33 75.56 13 1713 2.93 78.49 SUBTOTAL 45,893 IMISSING 12,562 21.49 99.98 TOTAL 58,455 30 TABLE 111.2 Birth date of cows by year, absolute, relative and cumulative frequency Birth Absolute Relative Cumulative date frequency frequency frequency (year) (PCT) (PCT) 1 ‘ 1,245 2.12 2.12 2 11,329 2.27 4.40 3 1,418 2.42 6.82 4 1,147 2.96 8.79 5 1,647 2.81 11.60 6 2,316 3.96 15.57 7 2,438 4.17 19.74 8 2,870 4.90 24.65 9 2,984 5.10 29.75 10 3,333 5.70 35.45 11 ‘ 3,068 5.24 40.70 12 2,786 4.76 45.47 13 2,640 . 4.51 49.98 14 2,203 3.76 53.75 15 ‘ 1,516 2.59 56.35 16 1,138 1.94 58.29 17 863 1.47 59.77 18 592 1.01 60.78 19 99 0.16 60.95 20 2 0.03 60.95 SUBTOTAL 35,623 MISSING 22,821 39 .04 TOTAL . ' 58,455 100.00 31 TABLE 111.3 Birth date of cows by month, absolute, relative and cumulative frequency Birth Absolute Relative Cumulative date frequency frequency frequency; (month), (PCT) (PCT) 1 ” ’ 3,345 5.72 5.72 2 - 3,238 5.53 11.26 3 3,150 5.38 16.65 4 3,160 5.40 22.05 5 3,463 5.92 33.82 6 3,463 5.92 33.82 7 3,164 5.41 39.24 8 3,296 5.63 44.88 9 3,234 5.53 50.41 10 3,224 5.53 55.94 11 3,032 5.18 61.13 12 3,248 5.55 66.68 SUBTOTAL 38,973 MISSING 19,482 33.32 TOTAL 58,455 ' 100.00 32 TABLE 111.4 Freshening date by year, absolute, relative and cumulative frequency Freshening Absolute Relative Cumulative date frequency frequency frequency (year) . (PCT) (PCT) 1 79' 8.13 0.13 2 178 0.30 0.43 3 790 1.35 1.79 4 1,396 2.38 4.17 5 1,207 2.06 6.24 6 1,158 1.98 8.22 7 1,299 2.22 10.44 8 1,479 2.53 12.97 9 1,957 3.34 16.32 10 2,531 4.32 20.65 11 3,263 5.58 26.23 12 3,767 6.43 32.67 13 3,290 5.62 38.30 14 3,725 6.37 44.67 15 4,639 7.93 52.61 16 5,141 8.79 61.40 17 4,728 8.08 69.70 18 4,510 7.71 77.20 19 4,063 ~ 6.95 84.16 20 1,771 3.02 87.19 SUBTOTAL 50,971 MISSING 7,484 12.80 TOTAL 58,455 99.99 33 TABLE 111.5 Freshening date by month, absolute, relative and cumulative frequency Freshening Absolute Relative Cumulative date frequency frequency frequency (month) _ (PCT) (PCT) ‘1 ' 4,288 7.33 7.33 2 4,018 — 6.87 14.20 3 4,210 7.20 21.41 4 4,325 7.39 28.81 5 4,620 7.90 36.71 6 4,450 7.61 44.32 7 .4,434 7.58 51.91 8 4,343 7.42 59.34 9 4,276 7.31 66.65 10 4,173 7.13 73.79 11 4,336 7.41 81.21 12 4,487 7.67 88.88 SUBTOTAL 51,960 MISSING - 6,495 11.11 TOTAL 68,455 . 100.00 34 TABLE 111.6 Lactation length in months, number of observations, absolute, relative and cumulative frequency Lactation Absolute Relative Cumulative lenggh frequency frequency frequency (month) , , (PCT) (PCT) 1 I 88 0.15 0.15 2 431 0.73 0.88 3 568 0.97 1.85 4 612 1.04 2.90 5 721 1.23 4.13 6 1,012 1.73 5.87 7 1,737 2.97 8.84 8 2,694 4.60 13.45 9 ' 4,731 8.09 21.54 10 45,834 78.40 99.95 SUBTOTAL 58,428~ MISSING 27 0 . 04 TOTAL 58,455 100.00 35 TABLE 111.7 Condition of cows affecting record, absolute, relative and cumulative frequency Code Absolute Relative Cumulative frequency frequency frequency (PCT) (PCT) 0'(N0rma1) 35,146 60.12 60.12 1 (Sold) 4,258 7.28 » 67.40 2 (Mastitis) 164 0.28 67.68 3 (Inf. Disease) 39 0.06 - 67.75 4 (Died) 1,224 2.09 69.85 5 (111) 638 1.09 70.94 6 (Incomplete record) 16,621 28.43 99.37 7 (Abort) 178 0.30 99.68 8 (Transport) 186 0.31 99.99‘ SUBTOTAL 58.454 MISSING 1 0.00 TOTAL 58,455 100.00 36 TABLE 111.8 Description of geographical region Code Altitude Temperature Rainfall Relative (mts)* (°C ** (mm annual) humidity 1 2600 - 3400 10 - 14 750 - 1250 83 - 88 2 2250 - 2700 14 - 17 840—1100 73 - 82 3 2700 - 3100 ll - 13 800 - 1700 77 - 80 4 2500 - 2900 12 - 13 504 - 1400 75 - 9O *Meters above sea level. **Celsius degrees. TABLE 111.811 Distribution of cows by region, absolute, relative and cumulative frequency Code Absolute Relative Cumulative frequency frequency frequency (PCT) (PCT) 1 10,913 18.66 18.66 2 5,743 9.82 28.49 3 20.761 35.51 64.00 4 20,911 35.77 99.98 SUBTOTAL 58,328 MISSING 127 0.21 TOTAL 58.455- 100.00 37 TABLE 111.9 Type of cows, absolute, relative and cumulative frequency Code Absolute Relative Cumulative frequency» frequency frequency (PCT) (PCT) 1 (Pure) 8,458 14.46 14.46 2 (Grade) 42,350 72.44 86.91 SUBTOTAL 50,808 MISSING 7.647 13.08 TOTAL 58.455 100.00 TABLE 111.10 Body conformation, absolute, relative and cumulative frequency Code Absolute Relative Cumulative frequency freguency frequency (PCT) (PCT) 0 (No classification) 53,784 92.00 92.00 1 (Poor) 4 0.00 92.01 2 (Regular) 605 1.03 93.05 3 (Good) 2,045 3.49 96.54 4 (More than good) 1,591 2.72 99.27 5 (Very good) 382 0.65 99.92 6 (Excellent) 44 0.07 TOTAL I 58.455 100.00 38 TABLE 111.11 Lactation number, absolute, relative and cumulative frequency 0 Lactation .Absolute Relative Cumulative number frequency frequency frequency_ (PCT) (PCT) 1 _ j ‘. . 19,137 32.73 32.73 2 11,521 19.70 52.44 3 6,931 11.85 64.30 4 2,812 4.81 69.11 5 . 1,189 2.03 71.14 6 528 0.90 72.05 7 157 0.26 72.32 8 25 0.04 72.36 SUBTOTAL 42,290 MISSING 16,175 27.65 TOTAL 58,455 100.00 39 TABLE 111.12 Distribution of cows by region, absolute, relative and cumulative frequency for the edited data set Area Code Absolute Relative Cumulative altitude frequency frequency frequency (PCT) (PCT) 2600-3100* ' 1 * 130 16.6 16.6 2250-2700* 2 7 .9 17.5 2700-3100* 3 295 37.6 55.0 2500-2900* 4 353 45.0 100.0 TOTAL 785 100.0 *Meters above sea level. TABLE 111.13. 40 Freshening year group at first lactation, absolute, relative and cumulative frequency Year Group Code Absolute Relative Cumulative frequency frequency frequency (PCT) (PCT) First* 1 ' ' 60 7.6 7.6 Second** 2 335 42.7 50.3 Third*** 3 390 49.7 100.0 TOTAL 785 100.0 *1953-1957 **l958-1962 ***l963-1967 TABLE 111.14 Freshening month group at first lactation, absolute, relative and cumulative frequency Month Group Code Absolute Relative Cumulative ' frequency frequency frequency_ (PCT) (PCT) Mar.-Apr. 2 133 16.9 31.0 May-June 3 142 18.1 49.0 July-Aug. 4 133 16.9 66.0 Sept.-Oct. 5 119 15.2 81.0 Nov.-Dec. 6 148 18.9 100.0 TOTAL 785 100.0 TABLE 111.15 41 Freshening year group at second lactation, absolute, relative and cumulative frequency Year Group Code Absolute Relative Cumulative frequency frequency frequency (PCT) (PCT) First* 1 24- 3.1 3.1 Second** 2 228 29.0 32.1 Third*** 3 533 67.9 100.0 TOTAL 785 100.0 *1953-1957 **1958-l962 ***1963-l967 TABLE 111.16 Freshening month group at second lactation, absolute, relative and cumulative frequency Month Group Code Absolute Relative Cumulative ' frequency frequency frequency (PCT) (PCT) Jan.-Feb. l 121 15.4 15.4 May-June 3 129 16.4 48.2 July-Aug. 4 132 16.8 65.0 TOTAL 785 100.0 nlIIIIIIl-IIIIII‘l-llll“ [I‘ll-Ill 42 TABLE 111.17 Farm distribution, absolute, relative and cumulative frequency Farm Absolute Relative Cumulative frequency frequency frequency (PCT) (PCT) 3- - 25 3.2 . 3.2 18 12 1.5 4.7 19 22 2.8 7.5 20 30 3.8 11.3 21 13 1.7 13.0 22 28 3.6 16.6 63 7 .9 17.5 101 37 4.7 22.2 104 65 8.3 30.4 107 31 3.9 34.4 108 11 1.4 35.8 116 13 1.7 37.5 119 l .1 37.6 120 27 3.4 41.0 121 68 - 8.7 49.7 126 23 2.9 52.6 134 17 2.2 45.8 136 2 .3 55.0 151 35 4.5 59.5 153 64 8 2 67.6 154 48 ~ 6.1 73.8 155 74 9.4 83.2 158 3 .4 83.6 159 4 .5 84.1 162 3 .4 84.5 168 .43 5.5 89.9 170 45 5.7 95.7 172 11 1.4 97.1 174 15 1.9 99.0 175 8 1.0 100.0 TOTAL 785 100.0 43 TABLE 111.18 Sire distribution, absolute, relative and cumulative frequency 0 Sire 1.D. Absolute Relative Cumulative frequency_ frequency frequency (PCT) (PCT) 24 26 52 .9 1.0 .5 o O‘ 206 232 251 252 269 299 378 380 431 432 440 445 447 448 450 481 504 536 546 548 588 604 609 612 641 678 690 697 744 862 866 888 938 944 949 952 1069 1094 1146 1172 o D H 0 H P‘k‘ 0 H P‘P‘ P‘P‘ 0 0 0 0 0 O 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 \J£~hob)P‘S‘P‘UJ¢>UQUJIDO‘NDUINJCJUD¢L$~NJ\JUIO\P‘$~W>\INDGDBJNDC>\JC>UIUIC>QD¢~P‘¢~¢>u> P‘ P‘ h) earl 0 0 0 0 0 0 H o a o O o o o 00 o o o o O O o o o o o O o o o o O o o 0 o 0 0 o O o o wNOWO‘bmwOObwaWNmJ-‘mHmwOMO‘LflHUIJ-‘O‘OHL‘O‘WOUIl-‘b i" NU‘INNU‘UJO‘NO‘WWNUJ-‘NNQUOkaNNNbMNOhwmmomem-fiowwvamQ h‘k‘ 0 0 0 N w UIUINHHOOOSOCDNO‘O‘O‘kJ-‘NI-‘COONNO‘O‘UIUNN|-‘HCOQQO‘UIUIWUOWNH OiuJuJUJUJUJUJth)thJthJthJthDth>h>h‘h‘h‘k‘k‘k‘k‘k‘h‘h‘k‘h‘ 44 TABLE 111.18 (continued) Sire 1.D. Absolute Relative Cumulative frequency frequency frequency; (PCT) (PCT) 1239 5 .6 36.3 1259 7 .9 37.2 1284 5 .6 37.8 1321 14 1.8 39.6 1322 7 .9 40.5 1360 8 1.0 41.5 1408 4 .5 42.0 1409 2 .3 42.3 1479 7 .9 43.2 1494 21 2.7 45.9 1519 22 2.8 48.7 1520 41 5.2 53.9 1523 24 3.1 56.9 1538 5 .6 57.6 1562 2 .3 57.8 1582 18 2.3 60.1 1587 13 1.7 61.8 1597 10 1.3 63.1 1662 2 .3 63.3 1783 2 .3 63.6 1801 4 .5 64.1 1870 16 .0 66.1 1871 12 1.5 67.6 1881 4 .5 68.2 1969 4 .5 68.7 1993 38 4.8 73.5 2001 5 .6 74.1 2091 31 3.9 79.4 2092 37 4.7 84.1 2144 2 .3 84.3 2276 2 .3 84.6 2289 6 .8 85.4 2313 3 .4 85.7 2346 4 .5 86.2 2372 3 .4 86.6 2665 4 .5 87.1 2749 7 .9 88.0 2994 2 .3 88.3 153006 2 .3 88.5 153012 3 .4 88.9 153013 2 .3 89.2 153016 2 .3 89.4 153018 2 .3 89.7 45 TABLE 111.18 (continued) Sire 1.D. Absolute Relative Cumulative frequency frequency frequency_ (PCT) (PCT) 153073 14 1.8 91.5 153101 4 ' .5 92.0 153141 15 1.9 93.9 153152 3 .4 94.3 153194 3 .4 94.6 153251 2 .3 94.9 153252 9 1.1 96.1 153283 2 .3 96.3 153311 5 .6 96.9 153312 7 .9 97.8 153314 3 .4 98.2 153316 12 1.5 99.7 153344 ._:2 .3 100.0 TOTAL 785 100.0 46 111.2.1 Models Used I Three mixed models were used in this study. The elements of Mbdel (i) were fixed factors area, year group, season groups, and its random portion sire. This model was used to test two way interactions. (1) Y ijklm 1 k + (Ba)jk + e(ijklm) - M+a +Bj+ a +Dl+(aBlj+(a3)ik Where the subscripts: 1 (a8) (a3) (83) indicates the area (i a 1, 2. 3. 4) indicates the year group (J - 1, 2, 3) indicates the season group (R a l, 2, 3, 4, 5, 6) indicates the sire (l - 1,...101) ‘ indicates the observation record within the th sire, kth season, jth year, and ith area or herd. is a common effect for all Y is is is is a a fixed area effect fixed year group effect fixed season group effect random sire effect fixed area-year group interaction fixed area-season group interaction fixed year group-season group interaction residual error group , 47 The elements of Model (ii) were fixed factors area, year season group, and the random factor sire. (11)Yijk1m' M+ai+B_. +ak+131 +£(ijklm) _ Where the subscripts: i ' indicates the area (i - l, 2, 3, 4) j indicates the year group (j - 1, 2, 3) k indicates the season group (R = 1, 2, 3, 4, 5, 6) 1 indicates the sire (l a 1,...101) m indicates the observation record within the th sire, kth season, jth year, and ith area or herd. and M is a common effect for all Y a is a fixed area effect 8 is a fixed year group effect 8 is a fixed season group effect D is a random sire effect a is a residual error In order to obtain more accurate sire estimates model (iii) in which herd is used instead of the fixed factor area was applied. (111) Y 8M + a1+ Bj + 3 + D1 + 8 ijklm k (ijklm) Where the subscripts: 1 indicates herd (i a l...30) j indicates the year group (j - 1, 2, 3) k indicates the season group (R = 1, 2, 3, 4, 5, 6) 1 indicates the sire (1 - 1,...101) 48 m indicates the observation record within the 1th sire,kth season, jth year, and ith area or herd and M is a common effect for all Y a is a fixed herd effect 8 is a fixed year group effect a is a fixed season group effect D is a random sire effect a is a residual error With models ii and iii first and second lactation records ' were analyzed separately and factor estimates, percentages of the variance and BLUP estimates of sires were obtained using the following equation} 253%. z. . (M v v ‘1 v zx z A + A’ [l L; 3 Where I, is the vector of observed records 8' .is the vector of unknown fixed effects for area, year, season is a random effect 35 & ’2. are design matrices describing the contribution of fixed (X) and random (2) effects to the records and, §:1 identity matrix in which the variance ratioCreZ/d's2 was added to the sire portion of the diagonal (submatrix _2_' E). Adkinson (1974) using Ecuadorian data obtained values of heritability of 0.32:0.02, 0.48:0.03, and 0.68:0.03 for milk yield, fat yield and fat percentage, respectively. These values were used to calculate the Scalars: 11.500 for milk yield, 7.3333 for fat 49 yield and 4.8824 for fat percentage, which were added to the E' Z 4. submatrix according with the analyzed Y. CHAPTER IV RESULTS AND DISCUSSION IV . 1 Records Management The analyzed data set tell us about management conditions on dairy farms in Ecuador during the study period. Frequency distributions for each one of the variables present in the records were obtained. A discussion of these follows. 1V.l.1 ‘Agg The cow's age varies form one to thirteen years. Twenty two 2 of the records had no age information. Sixty-four Z of the cow's were under 7 years and 362 were more than 7 years old (Table 111.1). Age is important in relation to genetic progress, generation interval and profitability obtained from an animal. Age at first freshening was 35.47 months and 67 months at second freshening. Roman (1970) found an age average of 3.22 months in registered and 33.66 months in grade cows at first freshening and 47.83 and 48.85 months for registered and grade cows, respectively, at second freshening. Longevity is possitively' correlated. with production traits to select sires in Ecuador. IV.1.2 Date of Birth In the study period, 392 of the records had no year of birth information and 33% were incomplete with regard to month 50 51 of birth (Tables 111.2 and 111.3). The difference of 67. between these two parameters suggest inaccurate record keeping at the farm. These factors are of primary importance to standardize milk production to mature equivalent. 1V.l.3 ‘Freshening;pate In the data set shown in Tables III.4 and 111.5, 132 of the records of year of freshening and 11% of the records of month of freshening were missing. Again the lost information is important in genetic type of work. Errors are always additive. IV.1.4 Lactation Lenth. For lactation by month (Table III.6) 27 records were missing and 78% of the records had complete. lactations. The incomplete lactation records increased from 0.22 in the first to 9% at the 9th month. Four 2 of the lactations were finished before the 3rd month period in which the lactation peak is achieved (Ferris, 1982). Other causes such as environmental, genetic, management or combinations of these could be the cause of incomplete lactations. IV.1.5 Condition Affecting the Record (CAR) One out of 58,455 records were missing in this variable (Table 111.7). Sixty Z of the records were normal lactations, 287. were incomplete lactations and 27. were reported from milking cows with mastitis or that had aborted. IV.1.6‘A£gg The analyzed data set is coming from the Andean Zone where dairy operations are principally located in valleys within 52 the different basins formed by the west and east mountain chains and their numerous connections between them. Four areas have been identified (Table III.8)-; their altitude goes from 2,500 to 3,400 meters above sea level; the weather is sui generis. The average temperature is between 10 to 14°C normally, but there are rare days with 1°C or highs of 29°C. Rainfall goes from 750 to 1,400 mm per /" year. It is not possible to define seasons as in the temperate areas of the world. However, two seasonal periods, wet and dry, are easily determined. Relative humidity varies from 73 to 902 the year around. Areas coded as 3 and 4 had the heavier cow density with 36% of the population (Table III.8.1). 1V.l.7 Type of Animal and Body Conformation The data set was made up of two types of animals: grade cows which formed the larger group with 732, and pure registered cows with 142. Thirteen Z of the records were missing (Table 111.9). Ninety two Z of the records had no type .classification. This high percentage indicates the number of grade cows. Eight 1 were classified as regular, good, more than good or excellent (Table 111.10). Body conformation is negatively correlated with production traits (Everett, 1977). Therefore, in Ecuador more emphasis must be given to age in sire selection. IV.1.8 -Lactation Number Records up to eight lactations (Table 111.11) have been recorded. Thirty three 2 corresponded to first, 202 to second, 122 to third lactation and the last 77. to lactations from the 53 fourth to the eighth. Twenty eight Z of the records were missing. 1V.l.9 Milk Production Actual 2X, 305 days of milk and fat yields and fat percentages were obtained from 8,168 records which constituted 432 of the. total first lactations, and from 6,770 records which constituted 582 of the total. second lactations (Table 1V.1). Production was 6,835 lb of milk with 3.44 fat 2 and 234 1b of fat in the first lactation; and 7,785 lb of milk with 3.41 fat 2 and 266 lb of fat in the second lactation. These values are lower than those found by Adkinson (1972) working with Ecuadorian data. After standardization to M.E. the number of records was reduced (Table 1V.2) to 4,932 records of first lactation and to 5,119 records of second lactation; the yields were 8,212 lb of milk, 3.35 fat 2 and 276 1b of fat in the first lactation and 8,654 lb of milk with 3.37 fat 2 and 291 lb of fat in the second lactation. 1V.2 ‘Statistical models From the 58,455 records only 1,848-records had complete and valuable information corresponding to 5% of the first and 82 of the second lactations. From these a: total of 1,570 records from both first and. second lactations were selected for analysis. The absolute, relative and cumulative frequencies from each one of the chosen variables for the statistical models used are shown in Tables 111.12 through 111.18. Table IV.3 shows R2 values resulting from fitting different models and "Y" variables in first and second 54 lactations. Mbdel (i) was used to determine the effect of fixed factors and its interactions. Both Models (ii) and (iii) were used to rank sires having as fixed and random factors area, year, season, and sire .or herd, year, season, and sire, respectively. Thesemodels were used because a confounding effect obtained with a model which included both area and herd effects. Herd effect was nested within area effect making unfeasible herd absorption. Analysis of variance for each one of the models is listed in Tables IV.4 to IV.8. IV.2.1 Examination of Factor Effects IV.2.1.l .éEEi Area was found as a highly significant factor (P<0.01) using Fisher's test in models (1) and (ii) for milk yield, fat Z and fat yield in both first and second lactation records (Tables IV.4 and IV.5). IV.2.1.3 Year GrOup Year group was highly significant (P <0.01) for milk yield, fat Z and fat yield in first lactation records (Table IV.5) and for milk yield and fat yield in second lactation records using model (ii) (Table IV.6). Fat Z using model (iii) (Table IV.8) as well as fat yield was significant (P<0.05) (Table IV.8). Similar results were obtained by Roman (1970), Adkinson (1972) and Rodriguez (1974). IV.2.1.4 Season Group Season-group was not a significant factor in any model for milk yield, fat Z or fat yield. Roman (1970) found 55 TABLE IV. 1 Milk and fat yield, in pounds and fat percentage (actual records) Mean (E) Std. Error Std. Dev. Range First Lactation (8168 Records) Milk yield (lb) 6835.45 1.90 172.35 1383 Fat (Z) 34.45 0.03 2.66 53 Fat yield (lb) 234.11 0.62 56.23 727 Second Lactation (6770 Records) Milk yield (1b) 7785.24 2.28 188.14 1377 Fat (Z) 34.16 0.03 2.69 51 Fat yield (lb) 1266.56 0.74 61.00 760 56 TABLE 1V.2 M11k.and fat yield, in pounds and fat percentage (mature equivalent records) Mean (i) Std. Error Std. Dev. Rangg_ First Lactation (4932 Records) Milk yield (lb) 8211.84 29.44 2068.07 15609.05 Fat (Z) - 3.35 0.00 0.31 5.16 Fat yield (lb) 275.85 0.94 66.36 894.08 Second Lactation (5119 Records) Milk yield (1b) 8654.17 28.53 2041.53 14047.70 Fat (Z) 3.37 0.00 0.27 5.05 Fat yield (lb) 291.22 0.90 64.90 728.46 57 TABLE IV.3 R2 statistics for the different models and "Y" variables in first and second lactations First Lactation Second Lactation Milk Yield Fat Z Fat Yield Milk Yield Fat Z Fat Yield Model 1 ' 0.3199 Model ii 0.2935 0.1688 0.3080 _ 0.2822 0.1279 0.2910 Model iii 0.4989 0.2123 0.4946 0.4824 0.1672 0.4762 TABLE IV.4 Summary of analysis of variance in first lactation region model (1) Source of df Mean Square F Test* variance Area 3 . 1.284 E8 - 43.34‘1) Year 2 3.215 E7 10.85(1) Season 5 2.648 E6 0.89 Sire 101 3.668 26 1.24(2) Area-year 4 5.761 E6 1.96(2) Area-season 13 2.095 E6 0.71 Year-season 10 2.436 E6 0.82 Error 646 2.962 E6 Total 784 *Fisher's variance ratio. (1) P<0.001. (2) P<0.10. 58 TABLE IV.5 Summary of analysis of variance in first lactation region model (iii) Source of df Mean Squares* variance Milk Yield Fat Z Fat Yield Area 3 1.284 E8(1) 0.55724‘1) 108303 Year 2 3.215 E7(l) 0.35379‘1) 42684‘1) Season 5 2.648 E6 0.04646 3366 Sire 101 3.668 E6(l) 0.02213 5066(1) Error 673 2.976 E6 0.03981 3134 Total 784 *Fisher's variance ration; test of significance. (1) P<0.00l. TABLE IV.6 Summary of analysis of variance in second lactation region model (ii) Source of df ’ Mean Sguares* Milk Yield Fat 2 Fat Yield Area 3 1.381 88(1) 0.48136 116437<1> Year 2 3.292 E7(l) 0.04700 38424<1> Season 5 4.005 E6 0.00914 3184 Sire 101 3.935 E6 0.04496 6192(1) Error 673 3.394 E6 0.03165 3499 Total 784 *Fisher's variance ratio; test of significance. (1) P<0.001. 59 TABLE IV.7 Summary of analysis of variance in first lactation herd model (iii) Source of df Mean Squares* variation Milk Yield Fat Z Fat Yield Herd 29 4.536 87(1) 0.13946(1) 45888 Year 2 2.248 E6 0.41402 8651 Season 5 2.016 E6 0.04475 2258 Sire 101 1.156 E6 0.02458 1526 Error 647 8.947 E5 0.03171 1506 Total 784 *Fisher's variance ratio, test of significance. (1) P<0.00l. 'TABLE IV.8 Summary of analysis of variance in second lactation herd model (111) Source of df Mean Squares* ' Milk Yield Fat Z Fat Yield Herd 29 4.962 E7(1) 0.11271(1) 49533‘1) Year 2 2.309 E6 0.08346(1) 3663(2) Season 5 1.794 E6 0.04611 2493 Sire 101 1.566 E6 0.03122 2018 Error 647 8.368 E5 0.02765 1265 Total 784 *Fisher's variance ratio; test of significance. (1) P< 0.001. (2) P< 0.05. 60 highly significant effects for fat Z and fat yield using three month season groups. However, for sire ranking purposes under Ecuadorian conditions (rainy and dry seasons) the season-group effect sill remains to be demonstrated. IV.2.1.5 Interactions Effects of the two way interactions area-year, area-season and year-season were tested in model (i) (Table IV.4). They were not significant (P<0.10) and consequently they were eliminated from.further analysis. IV.2.1.6‘§i£g Sire effects were found highly significant (P<0.001) in model (ii) for milk and fat yield in first lactation heifers and for fat yield in second lactation cows (Tables IV.5 and IV.6). Also, sire effects were significant (P <0.10) in model (1) (Table IV.4) but not significant for any variable in model (iii) in both first and second lactation records (Tables IV.7 and IV.8). These values agree with Roman (1970). IV.3 Site's Rank Rank of sires using models (ii) and (iii) were obtained. Top and bottom sires were the same for first and second lactations in each model for milk and fat yields. For fat Z the best and the worst sires were different in both lactations and models. The range in model (ii) was higher for the analyzed production traits than in model (iii) (Table IV.9). The sire's rank obtained in models (ii) and (iii) for milk 61 yield were compared to the rank for the other traits in first and second lactations using Spearman's correlation of ranks (Table IV.10). Higher correlations (P<0.0001) were obtained in model (iii) between lactationsaone and two. For lactation two model (ii) and for lactations one and two model (iii), the correlations were significant. The rank of BLUP estimates in 305 day first and second lactations for milk yield, fat Z, and fat yield in both models are summarized in Tables IV.11 to 1V.16. Herd model produced a higher R2 in both lactation records for all production traits. Smaller error variance (6'e2) was obtained using this model. Then, a more accurate estimate of sire effects could be obtained. The use of both area and herd models could be recommended in Ecuador. A computer run including herd in the model after deletion of all negative sire estimates obtained using model (ii) will produce a more confident estimation of sires. 62 TABLE 1V.9 Range of BLUP estimates and its standard error in models (ii) and (iii) for milk yield, fat Z, and fat yield in both first and second lactations Model (ii), Model (iii) Sire Estimate Std. Error Sire Estimate Std. Error MY 1523 1700.53 303.76 153194 620.93 392.73 944 -777.75 415.89- 949 -589.06 394.50 L1 F2 251 0.13 0.61 251 0.12 0.06 269 - 0.15 0.06 153344 - 0.13 0.08 FY 1523 69.52 10.77 153194 29.47 15.38 944 - 35.75 15.52 949 - 28.88 15.51 MY 1523 1348.38 _327.93 153194 860.28 442.54 944 -894.10 444.76 604 -596.16 409.25 L2 FZ 1881 0.13 0.08 678 0.12 0.09 153344 — 0.12 0.08 1519 - 0.14 0.05 FY 1523 52.81 11.41 153194 37.51 16.33 944 - 41.97 16.44 604 - 27.48 15.63 63 wk no.9 Nu NA wa.o no.6 >2 cam: o¢.c w¢.o ma.o rm no.0 me.o mc.o nc.o Nm HA qo.o mq.c mm.o mm.o No.0 >2 oa.o cq.o ¢¢.o No.0 co.o oa.c wm om.o c¢.o cm.o om.o mn.o om.o cm.° Nb NA mm.o «v.6 mm.o n¢.o he.o om.o «a.c om.o r: mm.o No.9 ca.o «m.o 00.9 mm.o Hm.o om.o mm.o rm 2 mm Nb >2 rm Nm #2 Mm Nb r: N sowuouooa H :oquouooa moofiuouuofi vow mufimuu HH< who: N oofiuouooa H cowuwuoma oou< .aaaav coo Aaae aaoooa ooa ooaaoaooooo some a.oaaooam oa.>a mam