Effect of selective recording on estimates of heritability

Resumen

In this work the effect of selective recording from older animals on estimates of heritability is analysed. Heritability of litter size was estimated in a pig population where data were selectively recorded for the older sows. Estimates were obtained in different data sets derived from all the available information. Data set were compared in terms of predictive ability. Underestimated heritability values were obtained when all available data were used. A maternal trait was also simulated and missing data generated by deleting records from females without known parents. Different animal models, usually applied when there is no selection or selective breeding is carried out, were considered in the heritability estimation of this trait. Results showed that none of these models offered unbiased estimates. Only the models that took into account the effect on the residual mean and the genetic mean and variance removed an important part of this bias. However, to be applied, selection carried out should be known. It is not obvious how to overcome a selective recording problem when the selection process followed is unknown.