The computations required to provide the information necessary in herd management are simple. In most instances, all that may be necessary is the individual unadjusted production data to aid in determining nutritional requirements and herd summaries.
Data processing for genetic improvement, on the other hand, requires a considerable amount of computation. The methods used depend on the nature of the breeding programme and the information required before selection decisions can be made. This chapter will deal with the procedures involved in correcting individual records for environmental effects and on the methods of ranking individuals for selection purposes.
Environmental factors tend to obscure true genetic ability. Performance records should therefore be adjusted for known environmental effects to remove some of the environmental variation. The value of such adjustment depends on the magnitude of the environmental effects and the accuracy with which they can be estimated. The procedures to correct for environmental effects are discussed below.
The simplest procedure, which is also efficient, is to relate the individual's record to the average of that of other animals subject to similar influences. Considering sheep for example, the weaning weight of a male lamb born as single to a 2-year-old ewe should be compared with the mean of other animals in the same category. This is possible only in large populations. In small populations, the method suffers from high sampling errors. Correction factors are therefore commonly used to adjust for environmental effects.
In principle, correction factors are only valid for the populations from which they are estimated. Estimates made within herds are, however, imprecise because of the small amounts of data. Furthermore, the computational work involved is also heavy. In practice, therefore, correction factors are estimated from data of many herds in a region and these are then used in all herds in that environment. Correction factors estimated from simple averages of group means are liable to be biased because of disproportionate numbers of these groups among other environmental classes. For example, correction factors for type of birth in sheep obtained from simple averages of single and twin-born lambs will be biased because of unequal numbers of these two groups in other classes such as dam-age or sex of lamb. Calculation of unbiased estimates requires simultaneous consideration of all effects. This is done by the least squares procedure.
The least squares procedure disentangles the different kinds of environmental effects. The system used is the method of fitting constants described by Harvey (1960). Before data can be prepared for least squares analysis by computer, the various environmental factors to be considered should be included in a model. The model can include continuous variables, e.g. age at weaning, and discrete classes e.g. sex. Interactions, if appropriate, may be included but they increase the amount of computation. It may be possible on the basis of preliminary analysis to omit some of the classes.
There should be at least two classes for a least squares analysis. Before assigning subclasses within a class, an estimate of the number of animals in each subclass should be made and where possible small subclasses should be combined or redistributed if the procedure is biologically acceptable. The analysis is more efficient if there are fewer subclasses and more individuals per class.
Correction factors are used in two ways to adjust data. The first adjusts all observations to the fitted mean by subtracting the constant from the observed value. The other adjusts all observations in relation to a subclass which is chosen as the standard (needs no correction), e.g. in adjusting for weaning weight of beef calves, the weaning weight of male calves born to dams over 6 years of age may be chosen as the standard. The latter method is more realistic as the adjusted data can be compared to those from other sources.
Environmental effects can also be corrected for by using a combination of correction factors and within subclass deviations. In this procedure, the data are first adjusted for all effects except those of herd and year using standard correction factors. Adjustments for the latter are then made by expressing the adjusted records as within-herd-year deviations. In small herds, the animal's own record should be excluded in computing the herd average, as otherwise autocorrelation arises between the herd mean and the animal's own production.
The performance of an animal can be influenced by external and internal factors. External factors affect the whole herd or flock, e.g. herd, year, season, while internal factors affect the individual animal, e.g. weaning age.
Herd, season and year differences: The usual procedure used to remove herd, year and seasonal variations are the contemporary comparison and herdmate comparison procedures. Both procedures are based on deviations of the individual performance from the average performance of animals milking in the same herd at the same time. But, while the contemporary comparison uses only the first lactation records, the herdmate method uses age-corrected records of all the animals commencing lactation during the year and season under consideration. When the number of first lactation animals in a herd within a year and season is small, contemporary comparisons are inefficient. The herdmate method is therefore to be preferred when herd sizes are small. Different procedures are used to obtain herdmate averages. Sweden uses the rolling herd average of which is obtained by dividing the total milk produced in the last 12 months by the total number of cow-years. In Denmark, it is calculated as the average of all corrected 305-day lactation yields completed during the past 12 months.
Settled village dairy herds in the tropics are usually very small in size. In such situations, the concept of a herd as a unit may have to be abandoned and herds in a location grouped into units on the basis of similarity of breed, production level and management.
Age at calving: Adjustment for age effects on first lactation performance is done on the basis of the regression of yield on age. Lactation yields of older cows are adjusted on the basis of parity to either first lactation or mature equivalent basis. Another method of correcting such records is on age, ignoring parity. This may be suitable when year and parity effects are confounded as when a new herd is started with young cows.
Lactation length: The 305-day lactation yield represents the yield from calving to 305 days irrespective of the length of lactation. when cows of European breeds are dried off before 305 days for management reasons, short records are corrected to 305 days. Such procedures are, however, not valid for cattle in the tropics, where short lactations appear to have a genetic basis. Calf death can also result in short lactations in zebu breeds that have a strong maternal instinct. However, correction for calf mortality may result in an upward bias in animals that also have a low potential for lactation length.
Calving interval: Milk yield is influenced by both the preceding calving interval and by the number of days pregnant during the current lactation. Since the relationship between length of calving interval and milk yield is curvilinear, regression using the linear and quadratic terms should be used to estimate the correction factors.
Correction factors for weaning and post-weaning weights are commonly used in developed countries. The Meat and Livestock Commission in UK uses a system that combines correction factors and within-subclass deviations. The correction factors adjust only for age of dam (Allen and Kilkenney, 1980). In USA, however, standard correction factors are used to adjust for all the effects. The UK system avoids the errors arising from the use of standard correction factors but suffers from the difficulties associated with inadequate number of contemporaries in small herds.
Standard correction factors are, however, not available in most tropical countries. The following environmental factors should be considered in their computation.
Herd and year: These effects are similar to those in dairy cattle.
Season: Variations in forage availability caused by marked differences in rainfall during the dry and wet seasons have a profound influence on performance of beef cattle in the tropics. Weaning weight depends on the season of birth while post-weaning weight is related to the season in which an animal reaches a given age.
Age of dam: The effect of age of dam on weaning weights in zebu cattle in the tropics may be less pronounced than in temperate countries because of the late age at first calving (Mosi, 1980). Sacker et al. (1971a), however, observed maternal effects even on post-weaning growth of Red Poll x Boran crosses in Uganda.
Previous parous state: Calves raised by dams parous the previous year are generally lighter at weaning than calves from dams which were non-parous and thus rested the previous year. The classes into which animals should be grouped are parous, parous but not having raised a calf and non-parous.
Current calving interval: The pregnancy status of the cow during weaning can influence weaning weight. However, it is not practicable to adjust for this effect because of the difficulty in obtaining mating records.
Age at weaning or post-weaning: The usual procedure for adjusting for age variations is to standardize the weight to a constant age by a simple multiplicative procedure,
where Wc is the estimated weight at a constant age C and W is the weight when the animal's age is D. Age adjustment can also be made by including it as a covariate in the least squares analysis.
The procedures that should be used for dairy sheep and goats are similar to those for dairy cattle. For sheep and goats kept for meat and/or fibre, the procedures are similar to those for beef cattle. The additional factor that should be considered is the maternal effect due to the type of rearing, the classes being single, twin and twin raised as single. Similarly, age of dam effects may be considered by classifying ewes into 2, 3 and 4–6 year old ewes. Since this effect is most marked between 2-year-old and older ewes, separation of dams into these two categories is often adequate, particularly in small flocks.
An animal's performance adjusted for environmental effects provides the best estimate of its breeding value where other information is not available. For characters of low heritability, the accuracy of this estimate is low. Information on earlier performance of the animal and records of its relatives can, however, increase the accuracy of the estimate. Males should be evaluated with greater accuracy than females because each male transmits its genes to a larger proportion of the progeny than a female.
Bulls: Progeny testing procedures for evaluating bulls are based on the contemporary comparison method (Robertson and Rendel, 1950) in which daughters of a bull are compared with daughters of other bulls milking in the same herd at the same time. The difference between the daughter and contemporary performance is weighted by the heritability (which depends on the numbers of daughters and contemporaries) to get an estimate of the sire's breeding value. Modifications of this procedure such as the cumulative difference (CD), best linear unbiased production (BLUP) and modified contemporary comparison (MCC) are now used to eliminate the bias that may arise due to genetic trend. All these methods take into account genetic differences between herds and the genetic competition faced by individual sire's daughters. Such sophisticated procedures may not be necessary in most parts of the tropics where progeny testing is in its infancy. However, where a testing programme has been going on for some time, adjustments for genetic trend may be necessary. For example, Mahadevan et al. (1970) observed that in the Bodles Jamaica Hope herd, the progeny test of bulls estimated in the first two years was higher than that of the same bulls estimated 7–10 years later due to the greater competition faced by their daughters in later years.
An assumption implicit in progeny testing of sires using the contemporary comparison method is that the dams of the contemporaries are a random sample of the population. This may not hold in situations where crossbred (exotic x indigenous) populations are used in progeny testing. The crossbred dams may differ in the percentage of exotic blood and therefore in their breeding values. Sires' breeding values estimated as
A = 2 b (Y - HY) should be corrected to
A = 2 b (Y - HY) - 0.5 h2 (D - HD) where
Y = yield of daughters, HY = yield of contemporaries,
D = yield of dams of daughters and
HD = yield of dams of contemporaries.
Where AI coverage is not extensive and there is concurrent use of AI and natural service (NS), daughters of AI bulls may milk concurrently with those of NS sires. The latter records could also be used in evaluating AI bulls by including in the least squares model the records of daughters of each young AI bull as a separate subclass and grouping records of all NS bulls' daughters into one subclass.
Cows: The breeding values of cows chosen as dams of bulls should be estimated accurately. This is done by the cow index. The procedure adopted is to first express the cow's record as a deviation from the herd mean so as to eliminate the influence of level of production in the herd. This information is then combined with that of its relatives, particularly those of its paternal daughters (sire's breeding value) and dam's performance to increase the accuracy of estimation of the breeding value. The weighting factors to be used to combine this information are given by Johansson and Rendel (1968).
Cow indices are also used in culling of cows but the procedures used are less sophisticated than in culling dams of bulls, since the same degree of accuracy is not required. Most often, the cow record expressed as a deviation from the monthly average of records completed during the last 12 months is adequate.
Cows: Cows are evaluated on the weight of the weaned calf, in the same manner that dairy cattle are evaluated on milk yield. In the cow index system used by the Meat and Livestock Commission in UK, the superiority in terms of the calf weight over that of its contemporaries is calculated for each calf of a cow. The sum of the contemporary comparisons is then adjusted for the number of calves produced by the cow and the number of contemporaries, to give the weighted cow index. This is analogous to the “most probable producing ability” (MPPA) used in USA.
Reproductive potential of the cow and calf survival are of great economic significance in beef cattle production in the tropics. Wilkins (1973) therefore suggests an index that includes these traits. The trait measured is the weight of weaned calf produced per month of productive calving interval, the latter being defined as the interval between the birth of the last calf reared to weaning and the birth of the calf under study. Thus, a calving in which the calf dies before weaning is deemed not to have occurred. The cow index is then calculated by expressing the above measure as a percentage of all cows calving during the month.
Bulls and heifers: Bulls and heifers are usually ranked on post-weaning weight, which is taken at about 18 months of age. For large herds, the least squares analysis gives unbiased estimates adjusted for environmental effects. These can be expressed as deviations from the overall adjusted mean for ranking purposes.
Wilkins (1973) describes a procedure suitable for small herds where at least 2–3 bulls are reared per month. All bulls between 14 and 18 months of age are weighed monthly and a linear regression of weight on age is calculated. At each weighing, every bull is given a rating which is the percentage deviation of the bull's weight from the predicted mean weight at that age. The mean of the five ratings (obtained from the 14th to the 18th month) is the bull's index. In this system, the population of bulls under test changes monthly, older bulls going out and young ones coming in.
Heifers can also be evaluated by the above procedure. However, since heifer rankings need not be as accurate as those of bulls, evaluation may be made on the basis of a single weighing at 24 months.
The procedures used for meat or wool breeds of sheep are similar to those for beef cattle. The most important characteristic in flocks that lamb seasonally the number of lambs born at each lambing. In the New Zealand national recording scheme the deviation of each ewe is calculated each year from the mean of all ewes of her age, and her average deviation calculated by summing this and dividing of the number of matings. To make records of ewes of different ages comparable, its average deviation is multiplied by a factor which takes into account the number of matings (k) on which the average is based, namely k/1 + (k - 1) t, where t is the repeatability of number of lambs born. Then, if F is the ewe's fertility index,
(average deviation in lambs born from mean of contemporaries).
The average deviation is now calculated from the deviations in individual years; it can also be calculated (Turner, 1978a) as (PD - D) where D = n/k (n = observed number of lambs born over k joinings) and D = average number born to ewes of ages up to and including the dam's present age.
Evaluation of rams for fleece or body weight can be made on the rams themselves but, for reproductive rate, measurements on female relatives are needed. The dam's fertility index is the most useful since it does not involve lengthening the generation interval. Turner (1978a) used ½h2F to score rams, h2 being the heritability. As this score can be positive or negative, Turner made adjustments which resulted in a score of 100 for ewes which had produced an average of one lamb per mating, and values above and below this for better or poorer performances.
The evaluation procedures for dairy sheep and goats are basically similar to those for dairy cattle.