Random regression models

Abstract

Performance of breeding animals is frequently measured on successive intervals, like milk traits, growth, feed intake, and exterior. The traits collected at different time or different places are not independent. Their average as well as standard deviation changes over time, and the correlation depends on the distance among them. For such kind of data, random regression models (RRM) describing production functions combined with covariance functions accommodating time- (place-)dependent (co)variance components are the method of choice. The models are compared to lactation yield models, to test interval model, and fixed regression models based on literature review, and illustrated using data from Slovenian black-white dairy-cattle population. RRMs reflect better biological background of animal performance over time and allow selection on the overall production level as well as the shape of production curve. They are in favor of more flexible recording scheme and thus, cost reduction for data collection. Furthermore, the genetic evaluation can be performed before lactation is finished shortening the generation interval and can be done more frequently giving a useful tool for herd management.