Comparison of heritability estimates from daughter on dam regression with three models to account for production level of dam.

Three models were used to estimate heritabilities for milk yields at different production levels and for different years as twice the regression of daughter residual effects on dam residual effects. The denominator is the residual mean square for dams. The numerator is the difference between the residual term for sum of dam's and daughter's records and sum of residual terms for records of dams and daughters. Model 1 included sire of daughter and herd-year-season of daughters only. Model 2 included sire of daughter, herd-year-season of dam, and herd-year-season of daughter. Model 3 included sire of daughter and herd-year-season of dam and herd-year-season of daughter combination. The weighted mean estimates for each method were, respectively, .35, .38, .38 for milk production and .61, .67, .67 for fat test. Yearly time trends in heritability were slightly positive for both milk production and fat test. Standard errors of heritability estimates from model 1 were 40 to 50% smaller than those from models 2 and 3 due to the smaller number of effects in the model. Estimates for model 2 from low to high production levels averaged .30, .38, .38, and .42 for milk yield and .64, .68, .67, and .71 for fat test.