Showing papers in "Interbull Bulletin in 1999"

Effect of health disorders on culling in dairy cows: a review and a critical discussion

Abstract: Risk of culling consequent to the main health disorders occurring in the current production systems is reviewed Survival analysis including health disorders as time-dependent variables is considered to be the most appropriate method to assess their effects because they allow a better description of the exact follow-up of disease history Farmers preferentially consider health events in the current lactation and/or those occurring in early stages of lactation for making culling decisions The unfavourable direct effects on culling of dystocia and udder disorders (mastitis and teat injuries) are clearly demonstrated, whereas there are variations between studies on the association between metabolic and reproductive disorders and culling These variations may be due to differences in study designs, populations involved and methods Consequences, in terms of estimated effect of health disorders, of methodological choices (eg whether or not including in the models descrip- tors for milk yield and/or reproductive performance) are discussed Metabolic and reproductive dis- orders may act indirectly through a subsequent decrease in milk yield and reproductive performance The impact of health disorders on longevity is on average weak, compared to the impact of low milk yield potential and poor reproductive performance Herd characteristics (availability of heifers, quota, farmer's attitude towards risk and uncertainty) modify the risk for a cow to be culled for a given health disorder Aims of further studies could be (1) to interpret the meaning and to analyse the reliability of culling reasons information, (2) to evaluate the relative effect on culling of health dis- orders and performance (milk yield and reproduction) in different parities, (3) to investigate the role of components of the herd effect on the risk of culling dairy cow / culling / health disorders / risk factors / survival analysis

Complex models, more data: simpler programming?

Abstract: The complexity of models used for or considered for use in genetic evaluation is increasing. Examples of new models are test-day models in dairy cattle, growth models in beef cattle, and models with dominance or/and QTL effects. These models are usually linear but analyses of some traits may require nonlinear models, which are usually more complicated to write and test. More complicated models may require a larger data set. In the future we may expect new types of models that will be used to analyze even larger data sets. In order to support new models, the computer programs need to be upgradeable and therefore easy to understand or simple. Programs in a matrix language are usually simple but inefficient and cannot work with larger data sets. In order to support large data sets, programs need to be efficient, which usually means complicated and hard to modify. Traditionally, mixed model packages available in animal breeding were written with efficiency in mind. Although they are useful at a time when they are developed, they become outdated. For example, none of the packages available in 1994 (Misztal, 94) supported the now-popular random regressions. Some of these packages have been updated to include random regressions, but some may be too complicated to update. Two developments can lead to simple yet efficient programs: increase in computer power and programming languages with objectoriented features. Greater computer power allows avoidance of optimizations that would complicate programs. Better programming languages allow to express the same algorithms simpler but as efficient. 2. Software complexity and Optimization

Two years of experience with the French genetic evaluation of dairy bulls on production-adjusted longevity of their daughters

Abstract: Four runs of the routine genetic evaluation of French dairy bulls on production-adjusted length of productive life of their daughters have been released since June 1997. This paper describes the current model of evaluation and how and why it has changed over the past two years. The computation of reliabilities is explained in details and the use of an approximation of the heritability of length of productive life on the observed scale is justified. Main results for fixed effects and relationships of sire effects with other traits are presented. A recurrent problem is the existence of strong upward biases in the evaluation of top proven sires with young batches of second daughters. This phenomenon, combined with the absence of reliable evaluations for young sires, explain a certain lack of concern for the new evaluation on behalf of breeders. Ways of improvements are indicated.

Projection of partial lactation records and calculation of 305-day yields in the Republic of Ireland

Abstract: Standard lactation curves for 2160 contemporary groups were derived for projecting records in progress and short lactations to 305-day yields in the republic of Ireland Projection was based on the deviation of the last test day and previous lactation 305-day yields from the corresponding expected yields from the standard lactation curves Calving season and age influenced the shape of the standard curves while level of production mostly affected the scale Correlation between yields projected from lactations 51days or more in lenght with realised yields were over 09 but accuracy increase with lenght of the record in progress ___________________________________________________________________________

Application of survival analysis in breeding for longevity

Abstract: Genetic evaluation for the length of productive life (LPL) in livestock species requires specific methods because of following reasons: (i) some animals are still alive at the time of evaluation and their complete LPL is not known, which implies that records of such individuals must be treated as censored; (ii) effects influencing LPL do not act linearly and vary with time; and (iii) the distribution of LPL data is often unknown and extremely skewed. Models based on survival analysis provide appropriate statistical tools for the analysis of LPL. These models are based on a hazard function λ(t) which describes the risk of being culled at time t conditional upon survival to time t. The hazard function of an animal is modeled as a product of a baseline hazard function, λ0(t), describing the ageing process, and a vector containing explanatory variables that supposedly influence the culling process and may depend on time. Because the form of λ0(t) is not always known with certainty, λ0(t) can be left unspecified, as in a Cox model. However, with complex models and large amount of data typical for national genetic evaluation, it is computationally advantageous to assume a parametric, e.g., Weibull hazard function. Such a model can also be easily extended to include a random sire or animal effect and enable estimation of genetic parameters for LPL. Routine genetic evaluation for LPL based on a Weibull model has been implemented in many countries worldwide. This paper gives an example of implementation of survival analysis techniques in dairy populations in Switzerland.

Multiple trait reduced rank random regression test-day model for production traits

Abstract: At the present the random regression (RR) test-day (TD) model is referred to be the most precise model for estimation of production traits in dairy cattle. It accounts for environmental variation within lactation and allows better description of effects due to pregnancy and maturity. Genetic variance among different TD yields can be modeled and allows estimation of breeding values for persistency. Multiple trait analysis increases accuracy of estimated breeding values (EBV) of less frequently measured traits. Hence, a multiple trait RR TD model should reduce biases in EBV’s and increase accuracy. However, implementation of RR models into national dairy cattle evaluation systems is a challenge due to computational limitations and difficulties in estimation of (co)variance components.

Fast and flexible program for genetic evaluation in dairy cattle

Abstract: A general purpose iteration on data BLUP-program (MiX99) was developed. Its fast performance makes continuous evaluation feasible, even when using a multiple-trait random regression test-day model. Reduction in computing time is due to four developments: preconditioned conjugate gradient as solving algorithm, a new technique for iteration on data, data compression, and ordering of equations by animal families. Three random regression test-day models with 0.25, 7.28, and 18.1 million unknowns in the mixed model equations were solved to compare MiX99 with the former available software. Total computing time (wall clock time) for first, second, and third model was 0.2, 3.3, and 9.2 hours, respectively, whereas corresponding values using the former software were 1.0, 19.7, and 172.4 hours. Results emphasize the superiority of the new implemented methods, especially when complexity of the model increases. The high performance was not impaired by the generality of the program, which allows a wide range of models. 1 Presented at the international workshop on high performance computing and new statistical methods in dairy cattle breeding, Tuusula, Finland, March, 18-20, 1999

Optimisation of progeny testing schemes when functional traits play an important role in the total merit index

Abstract: The inclusion of functional traits in a total merit index is expected to have some consequences on the design of progeny testing schemes in dairy cattle. Due to the usually low heritabilities of functional traits more daughter records per sire seem to be necessary to improve the accuracy of the breeding values of these traits. A complex deterministic approach was used to optimize the design of progeny testing schemes for Austrian Simmental and Brown Swiss cattle. The total merit index for Simmental included dairy traits (fat and protein yield), beef traits (daily gain, dressing percentage, EUROP grading score) and functional traits (functional longevity, persistency, fertility, calving ease, stillbirth and somatic cell count); the Brown Swiss total merit index consisted of dairy and functional traits only. The proportion of recorded cows mated with test bulls (test capacity) and the number of test bulls were varied. Annual monetary genetic gain and discounted profit were used to evaluate the different progeny testing schemes. The inclusion of functional traits in a total merit index has a positive effect on the annual monetary genetic gain. The annual genetic gains of functional traits are either positive or at least the negative trend can be reduced. Extending the number of daughter records per test bull from 60 (current situation) up to around 100 leads to a small increase of annual monetary genetic gain for Simmental, whereas no positive effect can be found for Brown Swiss. Discounted profit can be considerably improved, especially for Brown Swiss.  2002 Elsevier Science B.V. All rights reserved.

Extension of survival analysis models to discrete measures of longevity

Abstract: Proportional hazards models most often assume that failure times are expressed on a continuous time scale. There are common situations in animal breeding when this is not the case, for example when longevity is expressed in years, number of parities, etc... As a result, some basic assumptions of most survival models are violated. For example, the Cox’ partial likelihood is no longer correct. This paper presents a more appropriate strategy, the “grouped data model” of Prentice and Gloeckler (1978) for the analysis of discrete data, that maintains the assumption of proportional hazards. A reparameterisation of the model underlines the main differences with the Cox model or with a parametric regression model, the Weibull model. It also allows an easy modification of existing programs to make them suitable for the analysis of discrete survival data. The « grouped data model » is compared with continuous models on simulated data sets. The robustness of the Weibull model and the inadequacy of the Cox model on discrete data are illustrated.

Methodology for deriving non-market trait values in animal breeding goals for sustainable production systems

Abstract: An important aspect of the term sustainability is that environmental, genetic diversity, ethical and social aspects should be accounted for in addition to short and long-term economic value. The need for long-term biologically, ecologically and sociologically sound breeding goals is emphasized, because animal breeding determined only by short-term market forces has lead to unwanted side effects. Hence, a procedure for defining animal breeding goals with ethical priorities and weighing of market and non-market values is suggested. Implementation of nonmarket as well as market economic trait values in the aggregate genotype as suggested may allow for breeding programs that contribute to sustainable production systems. Methods for estimating non-market values are considered. These methods include among others: contingent valuation, choice experiments, revealed preference methods, implicit pricing and expert assessments.

Genetic variation in disease traits and their relationships with survival in Danish dairy cattle

Abstract: Data from the Danish health recording system on clinical mastitis, feet and leg diseases, metabolic diseases, reproductive dieseases and stayability has been analysed for Danish Holsteins, Red Danish and Danish Jerseys. Traits analysed were the incidence of the four diseases in the period 10 to 100 days from calving in 1st lactation. Records on 128,842 Danish Holstein cows initiating a 1st lactation in the period 1990-1995 were included in the analysis. The diseases were analysed as separate traits in multivariate analysis. The incidence in the period 10 to 100 days from calving in 1st lactation for clinical mastitis, feet and leg, metabolic and reproductive diseases were 26.3%, 6.2%, 6.2% and 11.1%, respectively. Stayability was defined as the incidence of cows that completed a 305-day record in 2nd lactation. (Co)variance components were estimated using a bivariate REML method with a linear sire model even though all traits analysed were observed on a binary scale. Estimates for heritabilities on the observed scale for clinical mastitis, feet and leg, metabolic and reproduction diseases were 0.05, 0.01, 0.01 and 0.02, respectively. The genetic correlations between disease and survival through 2nd lactation were 0.52, 0.43, 0.17 and 0.18 for clinical mastitis, feet and leg, metabolic and reproduction diseases. Further analyses will be carried out to develop an index for other diseases than Clinical mastitis based on health recordings in the first 3 lactations.

Stochastic simulation of breeding schemes for total merit in dairy cattle

Abstract: Total economic merit of dairy cattle depends on several traits. Despite of this, most studies of dairy cattle breeding schemes have assumed a single-trait breeding goal, i.e. milk production. In this paper, dairy cattle breeding schemes assuming a multi-trait breeding goal were studied using stochastic simulation. The breeding schemes studied consisted of a nucleus with 1500 females and a breeding population of 18 500 females. Bulls were assumed to be tested in a production population, which were not simulated. However, progeny test results of bulls were generated by simulating daughter yield deviations. All 10 traits in the current Danish breeding goal were simulated. The schemes studied included open vs. closed nuclei, selection for total merit vs. milk production, use of early predictors for milk production and different daughter group sizes. All simulations covered a 25-year period. Genetic gains were from 14 to 25 EURO per cow per year. Highest gains were obtained when selection was for total merit and early predictors were used. Higher daughter group sizes resulted in slightly higher economic gain, and the distribution of the gain were changed so the genetic gain for milk production were lowered and the genetic gain for cost reducing traits increased. For different schemes inbreeding increased from 0.98% up to 1.50% per year. On average genetic variance of the breeding goal were reduced by 30-39% in year 25.

An indirect approach for the estimation of genetic correlations between longevity and other traits

Abstract: To include functional longevity in selection schemes, the knowledge of the genetic correlations between longevity and other traits in the selection objective is required. In France, as in several other countries, functional longevity is nowadays evaluated using survival analysis models whereas other traits are evaluated using classical linear models. Then, it seems difficult to estimate the genetic relationship between these two types of traits directly from the data. We tried to circumvent this difficulty by computing genetic correlations from the sires’ estimated breeding values, using the Multiple Trait Across Country Evaluation (MACE) framework. Here we present an example of the application of such an approach for the analysis of the relationship between functional longevity and type traits. This study throws light on the danger of this procedure when some assumptions are not checked beforehand. In particular, the existence of non zero residual variances or a choice of bulls with unreliable evaluations may lead to biased or fluctuating results.

Potential use of parameters of the milk flow curve for genetic evaluation of milkability

Abstract: Starting in October 1998, a new milkmeter (LactoCorder) for dairy recording was introduced in Bavaria, Germany. With the LactoCorder milk yield is recorded but information is also available on milk flow. Based on thresholds of the flow rate, parameters are defined, e.g., maximum flow rate and duration of maximum flow rate. The LactoCorder is used on farms participating in test plans with alternate milk recording and sampling (AT), and on farms with milk recorded twice a day and alternate sampling (ATM), respectively, resulting in approximately 65% of the recorded cows having results obtained with the LactoCorder. A research project was initiated to implement a genetic evaluation for milkability, making use of the detailed information now available on milk flow curves. Since genetic relationships between udder health and milkability are well known, ongoing research is also focusing on this topic. For these analyses, information on somatic cell count will be added. Parameters of the milk flow curve included in a genetic evaluation will have to be selected with regard to milkability as well as to udder health. In analyses fixed effects of herd test day, lactation, days in milk, age at calving, time of milking, and bimodality of the flow curve were found to be highly significant for most of the traits.

Topics that may deserve further attention in survival analysis applied to dairy cattle breeding: some suggestions

Abstract: For the past ten years, the need to account for specificities of survival data (censoring, skewed distributions, time-dependent covariates) has lead to the development of methods, models, programs and routine genetic evaluations that are becoming standards. In this paper, I try to present some potential directions for further research, concentrating on modelling. These directions include the computation of cow EBVs, the use of timedependent sire effects, the implementation of multivariate analyses, the definition of survival traits on a lactation basis and the investigation of culling components. Obviously, this list is not exhaustive and some topics may be perhaps premature. They should be considered as the basis for discussion and for projects of collaborations.

Phenotypic relationship between type and longevity in the Holstein breed

Abstract: The relationship between type traits and longevity was studied in the French Holstein breed using a survival analysis model where the phenotypic value or the estimated genetic value of the cow for each type trait was included as a risk factor. This was done separately for two subpopulations (registered and non registered herds) and with or without adjustment for production traits, i.e., considering true or functional longevity. For all herds, udder traits (and above all, udder depth) clearly influence length of productive life. There seems to be a more drastic voluntary culling on type traits in registered herds. The correction for production traits, as a way to approximate functional longevity, increases the importance of udder traits and decreases the weight of capacity traits. The same results were obtained when the phenotypic value of the cow for type was replaced by her EBV. The relationship between longevity and type traits is most often nonlinear, in particular for udder traits.

The survival kit: a tool for analysis of survival data

Abstract: The Survival Kit is a set of programs that allows analysis of data which are measuring time untilsome defined event. Nonlinear Cox and Weibull regression models are featured. The programs are specifically adapted to the needs of animal breeders who tend to use very large data sets and want to estimate random effects. The covariance structure between observations based on genetic relationship can also be included. The programs are useful for analysis of (fixed) effects on longevity or any other time variable as well as for genetic evaluations on a national scale. As mixed or random effects models (called frailty models in survival analysis terminology) are not provided by standard statistical software packages, the Survival Kit can also be useful to researchers outside of animal science.

Discussion of heritability of survival traits

Abstract: Different definitions of heritability of survival traits are available in the literature on different scales and in different models. In the Weibull frailty model, heritability has been defined on the log time scale and transformed to the original time scale using Taylor series expansions. In the semiparametric Cox frailty model, heritability has been defined on the log frailty scale. In this paper, having presented the result that heritabilities on any (linear or approximated linear) time scales are identical, these measures of heritability are discussed. In the semiparametric Cox frailty model, there is a direct link between heritability defined on frailty scale and heritability on time scale. In the Weibull frailty model, the heritability that has been reported on the original time scale appears dubious.

Estimation of lifespan breeding values in the UK and their relationship with health and fertility traits

Abstract: In the UK BVs for herd life are currently predicted from a bivariate BLUP analysis of lifespan, measured in lactations, and a phenotypic index of the type traits most closely related to it. The type traits included in the index are foot angle (rg=0.22), fore udder attachment (rg=0.63), udder depth (rg=0.14) and teat length (rg=0.44). Lifespan PTAs are combined with production PTAs in the UK’s Profitable Lifetime Index (£PLI= -0.03*PTAMilk+0.6*PTAFat+4.04*PTAProt+38*PTALifespan). The economic value of lifespan (£38 per lactation) was calculated using a dynamic programming model by improving survival from lactation 1 to 4 by 1%, through a reduction in involuntary culling. Most of the benefit comes from a higher average level of production, as the herd age structure is more mature, and from a lower replacement cost. Selection using £PLI is expected to have an annual economic benefit of 5.9% over selection on production only. Including lifespan in the index may already be beneficial in terms of slowing down the rate of increase in incidence of health disorders and decline in fertility when selection is for production alone, so further analysis concentrated on the associations between lifespan and more direct measures of health and fertility. Calving interval was considered a measure of fertility, as calving dates are generally well recorded in most herds. Body condition score is recorded as part of the linear type classification scheme operated by Holstein UK and Ireland. Genetic correlations were inferred from the regression of daughter performance on sire PTAs for lifespan and were equal to -0.44 with calving interval, -0.11 with condition score, -0.22 with mastitis and -0.27 with lactation average SCC records. Thus cows with lower SCC, lower condition score and shorter calving intervals last longer in the herd. Future development of £PLI will include traits such as fertility and mastitis, either measured directly or using correlated traits, such as linear type scores, condition scores etc.

MCMC based estimation of variance components in a very large dairy cattle data set

Abstract: Bayesian MCMC methods, like Gibbs sampling, are often considered specialised tools for small scale data analysis. However, due to certain computational differences with e.g. REML, these methods also apply well to large scale data analyses which are practically infeasible by REML. The main attraction for very large data sets is that the basic computational algorithm for Gibbs sampling looks like an Gauss-Seidel iterative BLUP scheme, which can indeed be applied well to very large data sets. To illustrate this case, heritabilities for milk production and milk components were estimated in Dutch Friesian (DF) dairy cattle, (being) upgraded to Holstein Friesian (HF). The data contained 1 122 088 lactations of 585 785 cows and a total of 684 512 animals. Multiple lactations were analysed using a repeatability model. Fixed effects included were herd-year-season-parity (120 509 classes), month of calving (12 classes), permanent environmental effects for the 585 785 cows and regressions on fraction heterosis and fraction recombination loss to allow for the effects of crossing with HF. As can be expected from using such large amounts of data, estimated genetic parameters have an astonishing high accuracy. Heritabilities for milk, fat and protein were 0.401 (min 0.394, max 0.407), 0.362 (0.350-0.370) and 0.356 (0.352-0.365). Repeatabilities found were between 0.60 and 0.64 and the regression on heterosis was 293 kg milk. Using only first lactations, heritability for milk was significantly higher (0.464) and a breed difference between HF and DF of 570 kg was found.

Relationship between type traits and longevity in Austrian Symmental cattle

Abstract: The economic value of type traits in cattle is often seen in their effect on improvement of functional stability of cows. The relationship of various type traits obtained from a linear scoring system with the length of productive life of dairy cows of the Austrian Simmental breed was examined. A Cox survival model was used including the time-dependent effects of year-season, stage of lactation, level of milk yield, fat percent and protein percent relative to herd mates and, alternatively, one of the time-independent effects of 24 type traits. Type traits a were corrected for effects like classifier, stage of lactation and herd and fit as regressions with a linear and quadratic component. The data set consisted of a total of 5360 individuals. Of the 24 type traits, 12 showed a significant positive linear effect on length of productive life, one trait, muscularity of the front part of the body showed a significant negative effect. The largest positive effect was found for udder score. Quadratic terms were significant 6 times, and the signs of the quadratic regression coefficients indicated in all instances that more extreme expressions of a type trait have a negative effect on longevity. Results on the effect of type traits on longevity have to be viewed with caution because in many instances it might not be functional deficiencies but instead voluntary culling by the farmer for type that reduces the lifetime of cows.

Survival analysis using random regression models

Abstract: Censoring of records is a problem in the estimation of breeding values for longevity, because breeding values are required earlier in life than realised longevity can be measured. In this study we investigate the use of random regression models to analyse survival data, because this method combines some of the advantage of a multi-trait approach and the more sophisticated survival analysis. Production records in lactation 1 to 5 were available on 6320 cows in the UK, all having had the opportunity to survive 5 lactations. The random regression model contained a fixed cubic polynomial for lactation number (1 to 4), a herd effect (n = 167), a quadratic regression on milk yield within herd, a quadratic regression on age at calving within herd, Holstein percentage and year-season of calving effect (n=66). The additive animal genetic effects were modelled using a orthogonal polynomial of order 3 with random coefficients, and the error term was fitted as a diagonal matrix with different uncorrelated variances in each lactation. Variance components from the full (i.e. uncensored) data set, were used to estimate breeding values for survival in each lactation from both uncensored and censored data. Random censoring was applied proportionally to 0.3 or 0.6 of the cows, equal proportions of these censored animals had their last, last two or last three lactations set to missing. Two different procedures were applied: censoring or not censoring first lactation information. In the uncensored data, estimates of the residual variances were 0.15, 0.17, 0.17 and 0.18, and heritabilities were 0.03, 0.07, 0.05 and 0.01 for culling probability at the end of lactation 1 to 4, respectively. Breeding values for lifespan (calculated from the survival breeding values) had a range of 2.8 to 4.5 lactations and a standard deviation of 0.18. Correlations between predicted breeding values for 60 bulls, each with more than 20 daughters, from the various analyses ranged from 0.84 to 0.97. It is concluded that random regression analysis might be an alternative procedure to analyse censored survival data.

Genetic analysis of true profit for Spanish dairy cattle

Abstract: In this study, 7738 cows with production, type, and productive life traits from 239 herds with information on milk, meat, food, housing, and labour costs were used. A lifetime net profit value was calculated for each cow. An average cow had the first calf at 28.6 months of age, 6149 kg of milk in the first lactation, a productive life of 1236 days, 3.28 lactations and a net profit of 410 euros per year of productive life. Heritability of true profit was equal to 0.091. Genetic correlations with production and longevity traits were moderately high, ranged between 0.76 to 0.83. Genetic correlations with type traits were low, around zero, ranged between -0.33 and 0.21. A high genetic correlation was achieved with milk per day of productive life (0.92). Net merit index was the best predictor of the breeding value of true profit, with only 32% as coefficient of determination. Protein yield alone explained a 30% of the true profit. The best prediction of the true profit was achieved with the net merit index, plus rear leg side view, udder cleft and body depth, with 40% of total variation.

Genetic evaluation for length of productive life of Holstein cattle in Germany

Abstract: Since 1996, genetic evaluation of length of productive life of Black-and-White and Red-and-White Holstein dairy cows is carried out for West Germany. The SURVIVAL KIT package was applied for the first time in 1998 on data of all German HF-cows (Black & White and Red & White). Length of productive life is modelled with a Weibull model considering year-season within herd, lactation number * stage, relative milk yield within herd * year-season, relative change of herd size within year and a combined random genetic effect of sire and maternal grandsire of the cow. Sire solutions are standardized to a mean of 100 and a genetic standard deviation of 12, relative risks and an approximation of the breeding value expressed in days are also published. Solutions of fixed effects were plausible but averages of sire solutions differed among regions. Therefore a test-run was made in which interactions of region with lactation number * stage and with relative milk yield within herd * year-season were modelled. Solutions of the fixed effects showed that relative milk yield within herd * year-season should be considered in interaction with region.

Breeding value for functional longevity in the Netherlands

Abstract: In August 1999, the first official index for longevity will be published in The Netherlands. This index will be estimated using both information on longevity of sires’ daughters (direct breeding value), and information on sires’s breeding values for functional traits (indirect breeding value). The index aims to be an overall measure for the functionality, or ability to delay involuntary culling, of the daughters of a bull. As more information on the longevity performance of daughters becomes available, the weighting on the direct breeding value will be heavier. The direct breeding value is predicted using survival analysis techniques as programmed in the “Survival Kit”. Data on cows is available through the milk recording and IR the interaction between parity and stage of lactation; the interaction between herd, year, and season; production compared to herdmates; change in herd size; lactation value of cow; genetic group of maternal granddam; maternal grandsire; and sire. The pedigree file contains sires and maternal grandsires of bulls. The heritability estimated on log-scale is 0.04. When a young sire has longevity information on 30 daughters, the reliability of the direct index only will be 0.43. Regression techniques and selection index theory were used to identify which breeding values for functional traits explain most variation in direct breeding values for functional longevity. Traits that are suitable for including in the indirect breeding value are rump angle, teat placement, udder depth, feet and legs, somatic cell count (on log-scale) and interval calving – first insemination. Adding the information on conformation (60 daughters), fertility traits and SCC (100 daughters) to the direct information on longevity (30 daughters) increases the reliability of the combined index to 0.55. As the relative weighting of each source of information is dependent on the number of daughters, it will be determined for each bull separately.

The structure of the residual error variance of test day milk yield in random regression models

Abstract: The effect of assuming a constant residual (measurement) error structure in the analysis of test day milk records with a random regression model was investigated. Additive genetic and permanent environmental covariance components were each modelled with a 3 order orthogonal polynomial regression while measurement error was allowed to be either constant in all or some stages of lactation, or to vary from stage to stage. Coefficients of the covariance functions and residual variance components were jointly estimated by restricted maximum likelihood using the DFREML software package. There were significant increases (P<0.05) in the log-likelihood which suggests improvement in the fit of the derived covariance function as the constraint on residual error variance was removed. There was, however, little difference in estimates of the other variance components. Residual error was high in early lactation and declined rapidly to a stable value in mid/late lactation, so with a constant error assumption, residual variance was underestimated and heritability overestimated in early lactation.

Genetic evaluation of herd life in Canada: current status and future outlook

Abstract: Official genetic evaluations for dairy sires have been published in Canada since the winter of 1996. Evaluations are available for five major breeds (Ayrshire, Brown Swiss, Guernsey, Holstein, and Jersey). The most recent evaluation for Holsteins included records from 1,652,884 daughters of 80,197 sires, of which 5900 were eligible for official publication. Published evaluations combine information from daughter survival (direct herd life) and conformation traits (indirect herd life). Direct genetic evaluations of herd life are based on a multiple-trait linear animal model, in which survival in each of the first three lactations are considered correlated traits. Factors related to fat and protein yield are included in the model, so functional herd life is the trait evaluated. Heritability of herd life is assumed to be 0.04 for each lactation. Evaluations for the three lactations are then combined into an overall evaluation for direct herd life based on their relative economic values. Evaluations for indirect herd life are based on an index of evaluations for composite traits for mammary system, feet and legs, rump, and body capacity. Sire evaluations for the direct and indirect measures of herd life are combined into an overall evaluation for herd life using the methodology employed by Interbull to combine genetic evaluations across countries. Published sire evaluations are expressed as predicted transmitting abilities for the number of lactations of functional survival of daughters, centered at a base of 3 lactations, and range from approximately 2.3 to 3.5. Herd life is included in Canada's Total Economic Value (TEV) Index with a weight of 4 relative to 10 for production and 1.5 for udder health. Future changes in the genetic evaluation for herd life may include the adoption of a survival analysis. A survival analysis would be more appropriate statistically than is the current linear model and would account properly for censored records. Also, adoption of a survival analysis may increase the consistency of international evaluation, because several European countries have recently adopted survival analyses for their domestic evaluations. However, the survival analysis would be more computationally demanding and would likely be limited to a sire model. In addition, the survival analysis model would require the assumption that survival is the same genetic trait across an animal’s lifetime. The feasibility of a survival analysis of herd life in Canada is being studied.

Genetic improvement of functional traits in cattle - GIFT

Abstract: The recognition of the importance of functional traits in cattle (health, fertility, metabolic stress, longevity; Groen et al., 1997), and the possible role of animal breeding in avoiding deterioration and possibly improving functional traits has led to research activities in many EU countries. Research focuses on tool development (unambiguous trait definition, reliable trait recording and proper evaluation procedures) and tool implementation (optimisation of breeding programmes including production and functional traits). The state of the art in these areas was reviewed at a workshop in Gembloux (INTERBULL bulletin no. 12, 1996). At this workshop it was decided to apply for EU subsidies to organise an international concerted action on genetic improvement of functional traits in cattle to stimulate an efficient use of resources for scientific work. The EU subsidies were granted and the 3-year concerted action was started January 1997. The objectives of the GIFT concerted action are: • To bring together researchers and people from breeding organisations from member countries of the EU, and other countries with major interests in cattle production, to develop concepts for breeding of functional traits by defining breeding goals and strategies to achieve the goals; • To enhance collaborative efforts for the further development of efficient recording systems and breeding value estimation procedures; • To stimulate the exchange of existing knowledge about the genetic evaluation of functional traits; • To develop recommendations for breeding programmes for functional traits.

Genetic analysis of longevity data in the UK: present practice and considerations for the future

Abstract: The current genetic evaluation for herd life in the UK is reviewed, and alternative methods to analyse (discrete) longevity data discussed. Lifespan scores, i.e. the number of lactations a cows has survived or is expected to survive, and a phenotypic index of three type traits are combined using a bivariate animal model to produce UK national EBV for herd life. Preliminary analyses using proportional hazard models suggest that correlations between sire EBV from using either linear or proportional hazard models are high, and that fixed effects may be modelled better using proportional hazard models. For the use of EBV for survival, it is important to consider the appropriate formulation of the breeding goal. It is argued that a phenotypic adjustment for milk yield is not appropriate if the survival trait in the breeding objective is not adjusted accordingly. The emphasis on herd life may be too large if it is assumed that the calculated EBV are for avoidance of involuntary culling, since the genetic correlation between EBV for survival and involuntary culling are unlikely to be unity.