ABSTRACT
The present study evaluated the multiple traits model, as opposite to repeatability model approach in estimating heritability, breeding values and genetic correlation for four continuous traits (represent four parities) using bias, Mean squared error (MSE), Akaike information criteria (AIC) and Bayesian information criteria (BIC). The simulated base population consisted of 20, 60 or 100 sires. Each sire mated to 50 females to produce 1000, 3000 or 5000 progenies. The variance components modified to simulate three levels of heritability (h2), 0.05, 0.25 and 0.5. Genetic correlations (GC) among the four studied traits were 0.3, 0.5, 0.9 and one and residual correlation was 0.2. Twenty replicates were generated for each of the 36 combinations (three levels of h2 * four levels of GC * three levels of number of animals). These data sets (36 set) divided into two scenarios. The first one (27 set) was simulated as multiple traits and the second scenario (nine sets) was simulated as repeated measures. Each set of data in each scenario analyzed by multiple traits model (MTM) and repeatability model (RM). Correlations between the true and estimated breeding values of studied trait(s) estimated for the two types of analysis. Then, bias, MSE, AIC and BIC for all estimated values calculated as measures for comparing models of estimation. The mean estimates of h2 resulted from MTM and RM were 0.27, and 0.20, respectively. When level of GC increases, the mean estimate of h2 increases and reaches to equal the estimate of h2 resulted from MTM when GC=1. The bias and MSE of MTM are less than those of RM. The smallest estimates of bias and MSE were noticed at GC = 0.9 and 1. The lowest AIC and BIC values were observed when fitting RM with data of the two scenarios. Therefore, the two criteria favored RM. Correlations between true and estimated breeding values of the four traits in MTM were slightly better than in RM. The effect of type of model was significant (P<0.01). In addition, significant effects between number of animals levels, heritability levels and GC levels were observed (P<0.01). This study indicated that, multiple traits analysis is more accurate than repeated measurements analysis in estimating h2 and breeding values as concluded from results of bias and MSE. AIC and BIC were not the suitable criteria for selecting the appropriate model under the circumstances of this study.