Influência da estrutura da matriz de covariâncias na classificação de reprodutores caprinos

Abstract

Were analised records of weight-age from 412 goats from Moxotó breed, originated from 151 females and 50 males, born between 1984 and 1988, from march to july, in Pernambuco state, northeastern Brazil, totalizing 1749 records. Was considered a mixed model with the factors: sex (S), year of birth (A), month of birth (M), parturition type (T), bristle type (P), parturition order (O), linear effect of relative umidity (U), linear effect of temperature (C), linear effect of pluviosity (H), random effect of male (R), random effect of female (F), linear (D) and quadratic (D2) time effect in days. The objective was to verify the influence of covariance matrix type between weigths on prediction of random effects and the ranking of male goats. Were utilized the follow matrix type compond simmetry (CS), heterogeneity compond simmetry (CSH), autoregressive of order one (AR(1)), heterogeneous autoregressive of order one (ARH(1)), autoregressive moving average of order one (ARMA(1,1)), heterogeneous Toeplitz (TOEPH), non structured with correlation (UNR), variance component (VC), and spatial power (SP(POW)). Change in covariance matrix type changed the predicted random model effects. The most adequate matrix was ARH(1), influential effects were male (p<0,05), female (p<0,01), sex (p<0,01), parturition type (p<0,01), and linear and quadratic effect of time (p<0,01).