Files
Abstract
The first study estimates variance components for heat stress (HS) for the first three lactations using test day models. Repeatability (MREP) and random regression (MRRM) models included a random regression on a function of temperature-humidity index. Regular genetic variance increased from the first to second parity but slightly declined in the third parity. Genetic variance of HS strongly increased with parity. With the MRRM, the variance of the HS effect was about half of that of the MREP. The second study compares different computing options of preconditioned conjugate gradient algorithms for a large scale genetic evaluation using multiple-traits test-day random-regression models accounting for HS. Preconditioners were: diagonal, block diagonals due to traits (BT), block diagonal due to traits and correlated effects (BTCORR), and BT with the random effects reparameterized for diagonal (co)variance matrices within traits (BTDIAG). When sufficient memory is available, BTCORR was the fastest and the simplest to implement.