Determination of the number of test day records is required to replace lactation model with random regression model?

With development of mixed models, lactation model (LM) was replaced with random regression model (RRM). However, the minimum number of required test day (TD) records per animal to replace LM with RRM is a challenge in genetic evaluations. In this study, 381,236 test day records of 44,117 first parity dairy cattle which collected by Animal Breeding Center of Iran were used from 2006 to 2016. Based on number of TD records per animal, cows were divided into five groups by restricting cows that presented at least 2, 4, 6, 8 or 10 test day records in the lactation. The rank correlation between predicted breeding values (EBV) for LM and RRM irrespective of number of TD records was relatively moderate (0.44) and the rank correlation between two models using at least 2 TD records was 0.71. When 10 top percent of cows were used for comparison of LM with RRM, the rank correlation decreased to 0.08. The correlation of EBV of cows with ≥2 TD records with cows with ≥10 records in RRM was 0.85. The mean of breeding values accuracy in RRM was 4% higher than LM. Overall, use of RRM has advantages over LM and it is suggested to use RRM with at least 2 TD records instead of LM for genetic evaluation of milk yield trait.