Selecting random regression models under different minimum number of test day records

Autor: Claudio Napolis Costa, Alessandro Haiduck Padilha, Darlene dos Santos Daltro, José Braccini Neto, Jaime Araujo Cobuci
Rok vydání: 2017
Předmět:
Zdroj: Livestock Science. 199:69-73
ISSN: 1871-1413
Popis: The objective of this study was to compare EBVs, reliability and genetic parameters in random regression models with Legendre polynomials using structures of data sets with different minimum number of test days in lactation. The original data base was edited in order to prepare for subsets by deleting cows that did not have at least 4, 6, 8 or 10 test day (TD) records in lactation. The original intervals between monthly TD were used. Random regression models with third (M3), fourth (M4) and fifth-order (M5) Legendre polynomial were used. The lowest values of AIC, BIC, −2LogL and RV was found in the models with highest Legendre polynomials orders within structure with 6, 8 and 10 TD and lowest in structure with 4 TD. The eigenvalues indicated models with lowest Legendre polynomial orders as M3 and M4 in all structures. Heritability on days in milk ranged from 0.24 to 0.48 for M3 and from 0.17 to 0.31 for M4 and M5. Spearman correlations of EBVs of bulls and cows between M3, M4 and M5 were higher than 0.99 in all structures. Average reliability of EBVs of a group of bulls in common was around 0.82, 0.82, 0.80 and 0.63 in structures with at least 4, 6, 8 and 10 TD, respectively. Results indicate M3 and M4 as sufficient for genetic evaluations in all data sets of Holstein cattle. Random regression models will have similar reliability and ranks of EBVs in data sets with a minimum of 4, 6 or 8 TD.
Databáze: OpenAIRE
Externí odkaz: