A total of 335 records of swine litters from 102 Large White females that farrowed between 1985 and 1996, born in the seasons of september to february and march to august, in the Recife city, in Pernambuco state (Brazil), were analysed. The objective was to compare regression equations to predict stillborn rates. Angular transformation (TA), a generalized linear model (GLM) with binomial distribution and logit link (BL), a Poisson GLM with log link (PL), a mixed GLM (GLMM) with binomial distribution and logit link (BLM) and a Poisson GLMM with log link were compared. Were considered the following sources of variation: average weight of piglets at birth in kilograms (P), age of the mother at birth in years (I), number of males (M) and females (F) born. Selection criteria for best model were prediction capacity and residual analysis. The best model was the GLM with binomial distribution with logit link with de variables age of the mother, number of males, and number of females.
generalized linear model; regression equation; sow; stillborn
