Estimation of Genetic Parameters Using Sampled Data from Populations Undergoing Selection

¹ Department of Animal Science, Michigan State University, East Lansing 48824-1225

In populations undergoing selection, genetic (co)variances may be altered in amounts dependent on selection intensity among parents and the mating structure. In order to estimate the genetic parameters of the unselected population, all information that led to the current population must be included in the analysis. This is often not possible due to missing information or computer limitations, and, therefore, only samples of data and pedigree information of recent generations are included in analysis, and simplified operational models are used. Biases in genetic parameters, which were estimated by multitrait derivative-free REML method, were investigated in different strategies of sampling data and pedigree. In dual purpose cattle, in which young bulls are selected for growth before being progeny tested for milk yield, heritabilities and additive genetic correlations were all unbiased when all data and all relationships were used in an animal model. Using only recent data but all relationships in an animal model also gave unbiased estimates of heritabilities. Using an animal model for growth but a sire model for milk with all data gave an unbiased estimate of heritability for milk. When only recent data were used, the heritability estimate for milk was biased downward. In single purpose dairy populations, sire models gave biased estimates of genetic parameters even when all data were included in the analysis. Treating sire effects on second crop of daughters as fixed did not overcome selection bias.

Key Words: estimation • genetic parameter • data sampling • genetic phase disequilibrium

Submitted on April 16, 1990
Accepted on March 19, 1991