Assessment of the accuracy of selected models used to estimate the heritability coefficient in a cattle population

The aim of the study was to examine the use of different effects in models for estimating gestation length in dairy cows in order to determine which effects result in the most accurate assessment. The analysis was conducted on a database of over 2,5 million Polish Holstein-Friesian cows subject to use value assessment in Poland that calved between 2005 and 2010. Nine mixed linear models were analysed. Analysis of variance was performed for each model. The Akaike Information Criterion and Bayesian Information Criterion were calculated for each model. The model which takes into account two random effects, sire and herd, obtained the highest score for the feature of gestation length in the population. Analysis of various models confirms that the model should be adapted to a given population, and that its usefulness must be verified.