Optimization of test day for milk yield recording and sire evaluation in Murrah buffaloes

In the present study, random regression models (RRM) were used to estimate genetic parameters for test‐day milk yield in Murrah buffaloes using Legendre polynomial function (LP), with the objective to find the best combination of “minimum test‐day model,” which would be essential and sufficient to evaluate the trait successfully. Data included for analysis were 10,615 first lactation monthly test‐day milk yield records (5th, 35th, 65th, …, 305th) from 965 Murrah buffaloes for the period 1975–2018. Cubic to octic‐order orthogonal polynomials with homogeneous residual variances were used for the estimation of genetic parameters. Random regression models with sixth‐order were selected based on goodness of fit criteria like lower AIC, BIC and residual variance. Heritability estimates ranged from 0.079 (TD6) to 0.21(TD10). For both ends of lactation, the additive genetic and environmental variances were higher and ranged from 0.21 ± 0.12 (TD6) to 0.85 ± 0.35 kg2 (TD1) and 3.74 ± 0.36 (TD11) to 1.36 ± 0.14 kg2 (TD9), respectively. Between adjacent test‐day records, genetic correlation estimates ranged from 0.09 ± 0.31 (TD1 and TD2) to 0.97 ± 0.03 (TD3 and TD4; TD4 and TD5), but values gradually declined as the distance between test days increased. Negative genetic correlations were also obtained between TD1 with TD3 to TD9, TD2 with TD9 and TD10, and TD3 with TD10. On the basis of genetic correlations, models with 5 and/or 6 test‐days combination were able to account for 86.1%–98.7% of variation along the lactation. Models with fourth and fifth‐order LP functions were considered to account for variance with combinations of 5 and/or 6 test‐day milk yields. The model with 6 test‐day combinations had a higher rank correlation (0.93) with model using 11 monthly test‐day milk yield records. On the basis of relative efficiency, the model with 6 monthly test day combinations with fifth‐order was more efficient (maximum 99%) than the model using 11 monthly test‐day milk yield records. Looking into the similar accuracy with the 11TD model, and the low resources requirement, we recommend the use of the “6 test‐day combination model” for sire evaluation. These models may help in reducing the cost and time for data recording of milk yield.