CLASSICAL AND BOOTSTRAP ESTIMATES OF HERITABILITY OF MILK-YIELD IN ZIMBABWEAN HOLSTEIN COWS

The h2 of first lactation milk yield in Zimbabwean grade Holstein cows was estimated as the regression of daughter on the dams' adjusted mil k yield, using least squares, least absolute value, bootstrap restrict ed least squares, and bootstrap restricted least absolute value estima tors. The estimators were evaluated for efficiency and robustness in e stimating h2. The estimated h2 were .272, .265, .283, and .290 by leas t squares, least absolute value, bootstrap restricted least squares, a nd bootstrap restricted least absolute value procedures, respectively. The classic 90% confidence intervals for h2 were .030 to .514 by leas t squares and .056 to .474 by least absolute value procedures. The boo tstrap restricted least squares 90% confidence interval for h2 was .00 0 to .542, and the bootstrap restricted least absolute value 90% confi dence interval was .000 to .615. The point estimates by the four proce dures were comparable, but the bootstrap confidence intervals were mor e conservative than their classic counterparts. The least absolute val ue estimator was the most efficient and the second best in robustness. The least squares estimator was second best in efficiency and worst i n robustness. The bootstrap restricted least absolute value estimator was the least efficient and the second best in robustness.