Inferring causal relationships among growth traits in Kermani sheep applying structural equation modeling

Data on early growth traits of 1574 Kermani lambs, the offspring of 874 dams and 122 sires, were recorded between 1992 to 2010 at the Breeding Station of Kermani sheep breed in Shahr-e Babk city, Kerman province, Iran. The traits included the birth weight (BW), weaning weight (WW) and weight at six months of age (6MW). Preliminary investigations demonstrated the decisive effects of direct and maternal genetic influences on the expression of these traits. The data were used to adopt three models of genetic analysis. The first model was the standard multivariate model (SMM), which does not consider the causal relationships among the traits. The second model was the fully recursive multivariate model (FRM), which assumes the existence of causal influences of BW on WW and on 6MW, and the influence of WW on 6MW. The third model was the temporal recursive model (TRM), in which BW causally impacts on WW, and WW on 6MW. The Bayesian approach, via Gibbs sampling, was used for genetic analysis. The adopted models were compared considering the deviance information criterion (DIC) and mean square error of prediction (MSE) as criteria for assessing the predictive ability of the models. The DIC values revealed the superiority of TRM over FRM and SMM. Under the investigated multivariate models, the values of MSE were similar for BW but those obtained for WW and 6MW were the lowest under TRM. The causal effect of BW on WW and that of WW on 6MW were statistically significant estimates of 1.10 kg and 0.70 kg, respectively. Furthermore, accounting for causal relationships,the early growth traits in Kermani sheep may have advantageous impacts on prediction of the breeding values and consequently the accuracy of lamb ranking for selective breeding purposes. Generally, significant causal relationships were detected among the early body weight traits in Kermani sheep, and the superiority of TRM over other models showed the necessity of considering such causal influences in the genetic evaluation of these growth-related characteristics.