Palacios-Espinosa, Alejandro, Domínguez-Viveros, Joel, Padrón-Quintero, Yamariz, Rodríguez Castro, Manuel, Alonso Rodríguez-Almeida, Felipe, Luis Espinoza-Villavicencio, José, and Ysac Ávila-Serrano, Narciso
Nonlinear models (NLM) have been used to evaluate dairy cattle lactation curves (LC), and nonlinear mixed models are increasingly applied. The most appropriate model for estimating LC components in Siboney cattle was identified by applying a NLM for LC to 240 d; comparing a non-mixed NML vs a mixed NML; and estimating peak production (PMAX), total production at 240 d (PTOTAL) and days to maximum production (DPMAX). The database consisted of 15,324 records for daily milk production (kg) corresponding to first lactation in 2,809 Siboney cows (calving date in 2000-2012) in 28 herds in Cuba. Five NLM were evaluated: Wood, Wiltmink, Cobby, Brody and Sikka. Two SAS NLMIXED analyses were run: ANA1, including only the random effect of residuals; and ANA2, including a regression coefficient random effect plus residuals. Selection of the best fitting model was done based on average prediction error; prediction error variance; the Durbin Watson statistic; the determination coefficient (R2); the Akaike information criterion (AIC); and the Bayesian information criterion (BIC). The mixed NLM (ANA2) had the best fit: a) its R2 was higher (9 to 12 %); b) residual variance was notably reduced; and c) the AIC and BIC criteria exhibited a better fit. The model with the overall best fit was the ANA2 WOD, with estimates of 7.71 kg for PMAX, 40 d for DPMAX and 1,653 kg for PTOTAL. [ABSTRACT FROM AUTHOR]