Assessment of Homogeneity vs. Heterogeneity of Residual Variance in Random Regression Test-Day Models in a Bayesian Analysis

Random Regression Model for Genetic Evaluation and Early Selection in the Iranian Holstein Population

Simple Summary

The objective of this study was to use a model to predict breeding values for sires and cows at an early stage of the first lactation of cows and progeny groups in the Iranian Holstein population to support the early selection of sires. Our results show that we can select sires according to their daughters’ early lactation performance before they finish first lactation. Cross-validation results show that early selection accuracy can be high, and such an early selection can decrease the generation interval and lead to an increased genetic gain in the Iranian Holstein population.

Abstract

The objective of this study was to use a model to predict breeding values for sires and cows at an early stage of the first lactation of cows and progeny groups in the Iranian Holstein population to enable the early selection of sires. An additional objective was to estimate genetic and phenotypic parameters associated with this model. The accuracy of predicted breeding values was investigated using cross-validation based on sequential genetic evaluations emulating yearly evaluation runs. The data consisted of 2,166,925 test-day records from 456,712 cows calving between 1990 and 2015. (Co)-variance components and breeding values were estimated using a random regression test-day model and the average information (AI) restricted maximum likelihood method (REML). Legendre polynomial functions of order three were chosen to fit the additive genetic and permanent environmental effects, and a homogeneous residual variance was assumed throughout lactation. The lowest heritability of daily milk yield was estimated to be just under 0.14 in early lactation, and the highest heritability of daily milk yield was estimated to be 0.18 in mid-lactation. Cross-validation showed a highly positive correlation of predicted breeding values between consecutive yearly evaluations for both cows and sires. Correlation between predicted breeding values based only on records of early lactation (5–90 days) and records including late lactation (181–305 days) were 0.77–0.87 for cows and 0.81–0.94 for sires. These results show that we can select sires according to their daughters’ early lactation information before they finish the first lactation. This can be used to decrease generation interval and to increase genetic gain in the Iranian Holstein population.

Genotype by climate zone interactions for fertility, somatic cell score, and production in Iranian Holsteins

The objective of this study was to investigate genetic variation and genotype by environment (G × E) interactions for fertility (including age at first calving and calving interval), somatic cell score (SCS), and milk production traits for Iranian Holsteins. Different environments were defined based on the climatic zones (cold, semi-cold, and moderate) and considering the herd location. Data were collected between 2003 and 2018 by the National Animal Breeding Center of Iran (Karaj). Variance and covariance components and genetic correlations were estimated using 2 different models, which were analyzed using Bayesian methods. For both models, performance of traits in each climate were considered as different traits. Fertility traits were analyzed using a trivariate model. Furthermore, SCS and production traits were analyzed using trivariate random regression models (records in different climate zones considered as different traits). For the fertility traits, the largest estimates of heritability were observed in cold climate. Fertility performance was always better in cold environment. Genetic correlations between climatic zones ranged from 0.85 to 0.94. For daily measurements of SCS and production traits, heritability ranged from 0.031 to 0.037 and 0.069 to 0.209, respectively. Genetic variances were the highest in the semi-cold and moderate climates for the SCS and production traits, respectively. Furthermore, across the studied climates, 305-d genetic correlation ranged from 0.756 to 0.884 for SCS and from 0.925 to 0.957 for the production traits. The structure of genetic correlation within each climate indicated a negative correlation between early and late lactation for SCS, especially in the cold climate and for milk production in the moderate climate. For fat percentage, in all climatic zones, the lowest genetic correlations were observed between early and mid-lactation. In addition, for protein production in the cold climate, a negative correlation was observed between early and late lactation. Results indicated heterogeneous variance components for all the studied traits across various climatic zones. Estimated genetic correlations for SCS revealed that the genetic expression of animals may vary by climatic zone. Results indicated the existence of G × E interaction due to the climatic condition, only for SCS. Therefore, in Iranian Holsteins, the effect of G × E interactions should not be neglected, especially for SCS, as different sires might be optimal for use in different climatic zones.

Using Random Regression Models to Genetically Evaluate Functional Longevity Traits in North American Angus Cattle

Simple Summary

Cattle longevity is usually defined as the duration of life of a cow from first calving to death. In addition to a longer lifespan, it is crucial that cows are productive throughout their lives. Incorporating optimal indicators of productive longevity in breeding schemes will directly improve the economic profitability of the beef cattle herd and long-term sustainability of the industry. Thus, the impact of different longevity indicators in the selection of North American Angus cattle was evaluated and optimal parameters were defined to perform the evaluations.

Abstract

This study aimed to propose novel longevity indicators by comparing genetic parameters for traditional (TL; i.e., the cow’s lifespan after the first calving) and functional (FL; i.e., how long the cow stayed in the herd while also calving; assuming no missing (FLa) or missing (FLb) records for unknown calving) longevity, considering different culling reasons (natural death, structural problems, disease, fertility, performance, and miscellaneous). Longevity definitions were evaluated from 2 to 15 years of age, using single- and multiple-trait Bayesian random regression models (RRM). The RRM fitting heterogenous residual variance and fourth order Legendre polynomials were considered as the optimal models for the majority of longevity indicators. The average heritability estimates over ages for FLb (from 0.08 to 0.25) were always higher than those for FLa (from 0.07 to 0.19), and higher or equal to the ones estimated for TL (from 0.07 to 0.23), considering the different culling reasons. The average genetic correlations estimated between ages were low to moderate (~0.40), for all longevity definitions and culling reasons. However, removing the extreme ages (i.e., 2 and >12 years) increased the average correlation between ages (from ~0.40 to >0.70). The genetic correlations estimated between culling reasons were low (0.12 and 0.20 on average, considering all ages and ages between 3 and 12 years old, respectively), indicating that longevity based on different culling reasons should be considered as different traits in the genetic evaluations. Higher average genetic correlations (estimated from 3 to 12 years old) were observed between TL and FLb (0.73) in comparison to TL and FLa (0.64), or FLa and FLb (0.65). Consequently, a higher average proportion of commonly-selected sires, for the top 1% sires, was also observed between TL and FLb (91.74%), compared to TL and FLa (59.68%), or FLa and FLb (61.01%). Higher prediction accuracies for the expected daughter performances (calculated based on the pedigree information) were obtained for FLb in comparison to TL and FLa. Our findings indicate that FLb is preferred for the genetic evaluation of longevity. In addition, it is recommended including multiple longevity traits based on different groups of culling reasons in a selection sub-index, as they are genetically-different traits. Genetic selection based on breeding values at the age of four years is expected to result in greater selection responses for increased longevity in North American Angus cattle.

Random regression model with heterogeneous residual variance reduces polynomial order fitted for permanent environmental effect but does not affect genetic parameters for milk production in Sahiwal cattle

The random regression test-day models can accelerate the genetic improvement of Sahiwal cattle as test-day milk yield models offer a faster, accurate and economical approach of genetic evaluation. First three lactation monthly test-day records of Sahiwal cows calved between 1961 and 2012 at ICAR-National Dairy Research Institute, Karnal were used to fit random regression model (RRM) with various order of legendre polynomial, and a constant (RRM-HOM) or heterogeneous residual variance (RRM-HET). For both RRM-HOM and RRM-HET third order legendre polynomial for modelling additive genetic effects were found best. There was reduction in order of fit for modelling permanent environmental effects due to assumption of heterogeneous residual variance, as legendre polynomial of sixth order for RRM-HOM and fourth or fifth order for RRM-HET was found to be best, for modelling the permanent environmental effect. First two eigenvalues of additive genetic random regression coefficient matrix explained more than 99% of the additive genetic variation, whereas four eigenvalues explained ~98% of the permanent environment variations. First eigenfunction from both the models did not show any large change along lactation, suggesting most variation can be explained by genes that act in same way during lactation. The heritability estimates were generally low but moderate for some test-day milk yields, and it ranged from 0.007 to 0.088 for first, 0.044 to 0.279 for second, and 0.089 to 0.129 for third lactation from RRM-HOM. The values of genetic correlation between test-day milk yields ranged from 0.06 to 0.99 for first, 0.77 to 0.99 for second, and 0.07 to 0.99 for third lactation, from RRM-HOM. The value of permanent environment correlation ranged from 0.30 to 0.98 for first, 0.07 to 0.99 for second, and 0.16 to 0.98 for third lactation. The genetic correlations between two adjacent test-days were high (≥0.90). RRM-HET also gave similar heritability and correlation estimates. The rank correlation between sire breeding values for first, second, and third lactation, estimated using two models were 0.98, 1.00, and 0.99, respectively, indicating there was no difference in the ranking of animals using two models. Thus the random regression model with lower order of polynomial for modelling additive genetic effect and higher order polynomial for modelling animal permanent environmental effect was found suitable for genetic evaluation and both RRM-HOM and RRM-HET can be used for modelling test-day milk yield and breeding value prediction in Sahiwal cattle.

Models for Estimating Genetic Parameters of Milk Production Traits Using Random Regression Models in Korean Holstein Cattle

The objectives of the study were to estimate genetic parameters for milk production traits of Holstein cattle using random regression models (RRMs), and to compare the goodness of fit of various RRMs with homogeneous and heterogeneous residual variances. A total of 126,980 test-day milk production records of the first parity Holstein cows between 2007 and 2014 from the Dairy Cattle Improvement Center of National Agricultural Cooperative Federation in South Korea were used. These records included milk yield (MILK), fat yield (FAT), protein yield (PROT), and solids-not-fat yield (SNF). The statistical models included random effects of genetic and permanent environments using Legendre polynomials (LP) of the third to fifth order (L3–L5), fixed effects of herd-test day, year-season at calving, and a fixed regression for the test-day record (third to fifth order). The residual variances in the models were either homogeneous (HOM) or heterogeneous (15 classes, HET15; 60 classes, HET60). A total of nine models (3 orders of polynomials×3 types of residual variance) including L3-HOM, L3-HET15, L3-HET60, L4-HOM, L4-HET15, L4-HET60, L5-HOM, L5-HET15, and L5-HET60 were compared using Akaike information criteria (AIC) and/or Schwarz Bayesian information criteria (BIC) statistics to identify the model(s) of best fit for their respective traits. The lowest BIC value was observed for the models L5-HET15 (MILK; PROT; SNF) and L4-HET15 (FAT), which fit the best. In general, the BIC values of HET15 models for a particular polynomial order was lower than that of the HET60 model in most cases. This implies that the orders of LP and types of residual variances affect the goodness of models. Also, the heterogeneity of residual variances should be considered for the test-day analysis. The heritability estimates of from the best fitted models ranged from 0.08 to 0.15 for MILK, 0.06 to 0.14 for FAT, 0.08 to 0.12 for PROT, and 0.07 to 0.13 for SNF according to days in milk of first lactation. Genetic variances for studied traits tended to decrease during the earlier stages of lactation, which were followed by increases in the middle and decreases further at the end of lactation. With regards to the fitness of the models and the differential genetic parameters across the lactation stages, we could estimate genetic parameters more accurately from RRMs than from lactation models. Therefore, we suggest using RRMs in place of lactation models to make national dairy cattle genetic evaluations for milk production traits in Korea.

Polynomial order selection in random regression models via penalizing adaptively the likelihood

Orthogonal Legendre polynomials (LP) are used to model the shape of additive genetic and permanent environmental effects in random regression models (RRM). Frequently, the Akaike (AIC) and the Bayesian (BIC) information criteria are employed to select LP order. However, it has been theoretically shown that neither AIC nor BIC is simultaneously optimal in terms of consistency and efficiency. Thus, the goal was to introduce a method, 'penalizing adaptively the likelihood' (PAL), as a criterion to select LP order in RRM. Four simulated data sets and real data (60,513 records, 6675 Colombian Holstein cows) were employed. Nested models were fitted to the data, and AIC, BIC and PAL were calculated for all of them. Results showed that PAL and BIC identified with probability of one the true LP order for the additive genetic and permanent environmental effects, but AIC tended to favour over parameterized models. Conversely, when the true model was unknown, PAL selected the best model with higher probability than AIC. In the latter case, BIC never favoured the best model. To summarize, PAL selected a correct model order regardless of whether the 'true' model was within the set of candidates.

Modeling heat stress effect on Holstein cows under hot and dry conditions: selection tools

Data from milk recording of Holstein-Friesian cows together with weather information from 2 regions in Southern Spain were used to define the models that can better describe heat stress response for production traits and somatic cell score (SCS). Two sets of analyses were performed, one aimed at defining the population phenotypic response and the other at studying the genetic components. The first involved 2,514,762 test-day records from up to 5 lactations of 128,112 cows. Two models, one fitting a comfort threshold for temperature and a slope of decay after the threshold, and the other a cubic Legendre polynomial (LP) model were tested. Average (TAVE) and maximum daily temperatures were alternatively considered as covariates. The LP model using TAVE as covariate showed the best goodness of fit for all traits. Estimated rates of decay from this model for production at 25 and 34°C were 36 and 170, 3.8 and 3.0, and 3.9 and 8.2g/d per degree Celsius for milk, fat, and protein yield, respectively. In the second set of analyses, a sample of 280,958 test-day records from first lactations of 29,114 cows was used. Random regression models including quadratic or cubic LP regressions (TEM_) on TAVE or a fixed threshold and an unknown slope (DUMMY), including or not cubic regressions on days in milk (DIM3_), were tested. For milk and SCS, the best models were the DIM3_ models. In contrast, for fat and protein yield, the best model was TEM3. The DIM3DUMMY models showed similar performance to DIM3TEM3. The estimated genetic correlations between the same trait under cold and hot temperatures (ρ) indicated the existence of a large genotype by environment interaction for fat (ρ=0.53 for model TEM3) and protein yield (ρ around 0.6 for DIM3TEM3) and for SCS (ρ=0.64 for model DIM3TEM3), and a small genotype by environment interaction for milk (ρ over 0.8). The eigendecomposition of the additive genetic covariance matrix from model TEM3 showed the existence of a dominant component, a constant term that is not affected by temperature, representing from 64% of the variation for SCS to 91% of the variation for milk. The second component, showing a flat pattern at intermediate temperatures and increasing or decreasing slopes for the extremes, gathered 15, 11, and 24% of the variation for fat and protein yield and SCS, respectively. This component could be further evaluated as a selection criterion for heat tolerance independently of the production level.

Time trends, environmental factors and genetic basis of semen traits collected in Holstein bulls under commercial conditions

The fact that results of artificial insemination (AI) are declining in highly selected dairy cattle populations has added a renewed interest to the evaluation of male fertility. Data from 42,348 ejaculates collected from 1990 to 2007 on 502 Holstein bulls were analysed in a Bayesian framework to provide estimates of the evolution of semen traits routinely collected in AI centres throughout the last decades of intense selection for production traits and estimate genetic parameters. The traits under consideration were volume (VOL), concentration (CONC), number of spermatozoa per ejaculate (NESPZ), mass motility score (MM), individual motility (IM), and post-thawing motility (PTM). The environmental factors studied were year-season and week of collection, which account for changes in environmental and technical conditions along time, age at collection, ejaculate order, time from previous collection (TPC) and time between collection and freezing (TCF) (only for PTM). Bull's inbreeding coefficient (Fi), bull's permanent environmental and additive genetic effects were also considered. The use of reduced models was evaluated using the Bayes factor. For all the systematic effects tested, strong or very strong evidence in favour of including the effect in the model was obtained, except for Fi for motility traits and TCF for PTM. No systematic time trends for environment or bull effects were observed, except for PTM, which showed an increasing environmental trend, associated with improvements in freezing-thawing protocols. Heritability estimates were moderate (0.16-0.22), except for IM, which presented a low value (0.07). Genetic correlations among motilities and between motilities and CONC were large and positive [0.38-0.87], VOL showed a negative correlation with CONC (-0.13) but with ample HPD 95%. The magnitude of heritabilities would allow an efficient selection if required and grants the use of these traits as indicators of the sperm viability component of bulls breeding soundness.

Estimation of genetic parameters for milk yield in Murrah buffaloes by Bayesian inference

Random regression models were used to estimate genetic parameters for test-day milk yield in Murrah buffaloes using Bayesian inference. Data comprised 17,935 test-day milk records from 1,433 buffaloes. Twelve models were tested using different combinations of third-, fourth-, fifth-, sixth-, and seventh-order orthogonal polynomials of weeks of lactation for additive genetic and permanent environmental effects. All models included the fixed effects of contemporary group, number of daily milkings and age of cow at calving as covariate (linear and quadratic effect). In addition, residual variances were considered to be heterogeneous with 6 classes of variance. Models were selected based on the residual mean square error, weighted average of residual variance estimates, and estimates of variance components, heritabilities, correlations, eigenvalues, and eigenfunctions. Results indicated that changes in the order of fit for additive genetic and permanent environmental random effects influenced the estimation of genetic parameters. Heritability estimates ranged from 0.19 to 0.31. Genetic correlation estimates were close to unity between adjacent test-day records, but decreased gradually as the interval between test-days increased. Results from mean squared error and weighted averages of residual variance estimates suggested that a model considering sixth- and seventh-order Legendre polynomials for additive and permanent environmental effects, respectively, and 6 classes for residual variances, provided the best fit. Nevertheless, this model presented the largest degree of complexity. A more parsimonious model, with fourth- and sixth-order polynomials, respectively, for these same effects, yielded very similar genetic parameter estimates. Therefore, this last model is recommended for routine applications.

Genetic analysis of longitudinal data in beef cattle: a review

Currently, many different data types are collected by beef cattle breed associations for the purpose of genetic evaluation. These data points are all biological characteristics of individual animals that can be measured multiple times over an animal's lifetime. Some traits can only be measured once on an individual animal, whereas others, such as the body weight of an animal as it grows, can be measured many times. Data such as growth has been often referred to as "longitudinal" or "infinite-dimensional" since it is theoretically possible to observe the trait an infinite number of times over the life span of a given individual. Analysis of such data is not without its challenges, and as a result many different methods have been or are beginning to be implemented in the genetic analysis of beef cattle data, each an improvement over its predecessor. These methods of analysis range from the classic repeated measures to the more contemporary suite of random regressions that use covariance functions or even splines as their base function. Each of the approaches has both strengths and weaknesses in the analysis of longitudinal data. Here we summarize past and current genetic evaluation technology for analyzing this type of data and review some emerging technologies beginning to be implemented in national cattle evaluation schemes, along with their potential implications for the beef industry.

Use of test-day records beyond three hundred five days for estimation of three hundred five-day breeding values for production traits and somatic cell score of Canadian Holsteins

The Canadian Test-Day Model includes test-day (TD) records from 5 to 305 d in milk (DIM). Because 60% of Canadian Holstein cows have at least one lactation longer than 305 d, a significant number of TD records beyond 305 DIM could be included in the genetic evaluation. The aim of this study was to investigate whether TD records beyond 305 DIM could be useful for estimation of 305-d estimated breeding value (EBV) for milk, fat, and protein yields and somatic cell score. Data were 48,638,184 TD milk, fat, and protein yields and somatic cell scores from the first 3 lactations of 2,826,456 Canadian Holstein cows. All production traits were preadjusted for the effect of pregnancy. Subsets of data were created for variance-component estimation by random sampling of 50 herds. Variance components were estimated using Gibbs sampling. Full data sets were used for estimation of breeding values. Three multiple-trait, multiple-lactation random regression models with TD records up to 305 DIM (M305), 335 DIM (M335), and 365 DIM (M365) were fitted. Two additional models (M305a and M305b) used TD records up to 305 DIM and variance components previously estimated by M335 and M365, respectively. The effects common to all models were fixed effects of herd x test-date and DIM class, fixed regression on DIM nested within region x age x season class, and random regressions for additive genetic and permanent environmental effects. Legendre polynomials of order 6 and 4 were fitted for fixed and random regressions, respectively. Rapid increase of additive genetic and permanent environmental variances at extremes of lactations was observed with all 3 models. The increase of additive genetic and permanent environmental variances was at earlier DIM with M305, resulting in greater variances at 305 DIM with M305 than with M335 and M365. Model M305 had the best ability to predict TD yields from 5 through 305 DIM and less error of prediction of 305-d EBV than M335 and M365. Model M335 had smaller change of 305-d EBV of bulls over the period of 7 yr than did M305 and M365. Model M305a had the least error of prediction and change of 305-d EBV from all models. Therefore, the use of TD records of Holstein cows from 5 through 305 DIM and variance components estimated using records up to 335 DIM is recommended for the Canadian Test-Day Model.

Exploring the Use of Random Regression Models with Legendre Polynomials to Analyze Measures of Volume of Ejaculate in Holstein Bulls

Artificial insemination centers routinely collect records of quantity and quality of semen of bulls throughout the animals' productive period. The goal of this paper was to explore the use of random regression models with orthogonal polynomials to analyze repeated measures of semen production of Spanish Holstein bulls. A total of 8,773 records of volume of first ejaculate (VFE) collected between 12 and 30 mo of age from 213 Spanish Holstein bulls was analyzed under alternative random regression models. Legendre polynomial functions of increasing order (0 to 6) were fitted to the average trajectory, additive genetic and permanent environmental effects. Age at collection and days in production were used as time variables. Heterogeneous and homogeneous residual variances were alternatively assumed. Analyses were carried out within a Bayesian framework. The logarithm of the marginal density and the cross-validation predictive ability of the data were used as model comparison criteria. Based on both criteria, age at collection as a time variable and heterogeneous residuals models are recommended to analyze changes of VFE over time. Both criteria indicated that fitting random curves for genetic and permanent environmental components as well as for the average trajector improved the quality of models. Furthermore, models with a higher order polynomial for the permanent environmental (5 to 6) than for the genetic components (4 to 5) and the average trajectory (2 to 3) tended to perform best. High-order polynomials were needed to accommodate the highly oscillating nature of the phenotypic values. Heritability and repeatability estimates, disregarding the extremes of the studied period, ranged from 0.15 to 0.35 and from 0.20 to 0.50, respectively, indicating that selection for VFE may be effective at any stage. Small differences among models were observed. Apart from the extremes, estimated correlations between ages decreased steadily from 0.9 and 0.4 for measures 1 mo apart to 0.4 and 0.2 for most distant measures for additive genetic and phenotypic components, respectively. Further investigation to account for environmental factors that may be responsible for the oscillating observations of VFE is needed.

Genetic Analysis of Milk Production Traits of Polish Black and White Cattle Using Large-Scale Random Regression Test-Day Models

Genetic parameters for milk, fat, and protein yield and persistency in the first 3 lactations of Polish Black and White cattle were estimated. A multiple-lactation model was applied with random herd-test-day effect, fixed regressions for herd-year and age-season of calving, and random regressions for the additive genetic and permanent environmental effects. Three data sets with slightly different edits on minimal number of days in milk and the size of herd-year class were used. Each subset included more than 0.5 million test-day records and more than 58,000 cows. Estimates of covariance components and genetic parameters for each trait were obtained by Bayesian methods using the Gibbs sampler. Due to the large size and a good structure of the data, no differences in estimates were found when additional criteria for record selection were applied. More than 95% of the genetic variance for all traits and lactations was explained by the first 2 principal components, which were associated with the mean yield and lactation persistency. Heritabilities of 305-d milk yield in the first 3 lactations (0.18, 0.16, 0.17) were lower than those for fat (0.12, 0.11, 0.12) and protein (0.13, 0.14, 0.15). Estimates of daily heritabilities increased in general with days in milk for all traits and lactations, with no apparent abnormalities at the beginning or end of lactation. Genetic correlations between yields in different lactations ranged from 0.74 (fat yield in lactations 1 and 3) to 0.89 (milk yield in lactations 2 and 3). Persistency of lactation was defined as the linear regression coefficient of the lactation curve. Heritability of persistency increased with lactation number for all traits and genetic correlations between persistency in different lactations were smaller than those for 305d yield. Persistency was not genetically correlated with the total yield in lactation.

Comparison of random regression test-day models for Polish Black and White cattle

Test-day milk yields of first-lactation Black and White cows were used to select the model for routine genetic evaluation of dairy cattle in Poland. The population of Polish Black and White cows is characterized by small herd size, low level of production, and relatively early peak of lactation. Several random regression models for first-lactation milk yield were initially compared using the "percentage of squared bias" criterion and the correlations between true and predicted breeding values. Models with random herd-test-date effects, fixed age-season and herd-year curves, and random additive genetic and permanent environmental curves (Legendre polynomials of different orders were used for all regressions) were chosen for further studies. Additional comparisons included analyses of the residuals and shapes of variance curves in days in milk. The low production level and early peak of lactation of the breed required the use of Legendre polynomials of order 5 to describe age-season lactation curves. For the other curves, Legendre polynomials of order 3 satisfactorily described daily milk yield variation. Fitting third-order polynomials for the permanent environmental effect made it possible to adequately account for heterogeneous residual variance at different stages of lactation.

Comparing alternative definitions of the contemporary group effect in Avileña Negra Ibérica beef cattle using classical and Bayesian criteria1

Data on weaning weight from 12,740 animals were used to compare different definitions of contemporary groups (CG) for the genetic evaluation of the Avilena Negra Iberica beef cattle breed. Six alternative definitions for the CG effect were considered: herd-year-season of calving (HYS), with seasons defined according to the four natural seasons; herd-year-month of calving (HYM); herd clusters of 30 d (HC30-30) or 90 d (HC90-90); and adaptive herd clusters with two time limits, 30 and 90 d (HC30-90), and 30 and 180 d (HC30-180). A minimum of five observations in each CG class was required. This rendered substantial differences in loss of information, ranging from 0.7% of the total number of records for HC30-180 to 14% for HYM. Several classical statistics and Bayesian criteria for statistical model comparison were used. The use of classical criteria, such as the between- and within-CG variation and the accuracy of prediction, can be controversial because of their dependency on the unknown variance components. Residual variance decreased with the decrease in time span associated with the definition of CG. This was expected in this population because environmental conditions are highly variable throughout the year. However, estimates of the additive genetic variance for direct effects, which should not be affected by the definition of CG, were substantially larger for definitions involving larger time periods (HYS, HC90-90). When parameters used in the current evaluation procedure were used with all data sets, CG involving 30 d (HYM and HC30-30) were optimal in terms of providing the lowest/largest within-/between-CG variation. On the other hand, CG involving 90 d (HYS and HC90-90) yielded the poorest within-/between CG variation, with only a slight improvement of accuracy of prediction of direct genetic values over the other definitions. Bayes factors and cross-validation predictive densities allowed for improved discrimination among models. Models including CG spanning 30 d were more plausible and showed better predicting ability than models spanning 90 d. Adaptive CG showed intermediate results. Overall, it seems that average time span rendered by the different definitions had a major effect on the ranking of models. However, from the breeder's point of view, the loss of information associated with definitions involving shorter periods of time, such as HYM or HC30-30, might be unacceptable.