Abstract
The derivation and comparison of multiple regression equations for estimating solids-not-fat (SNF) using specific gravity and percentage milk fat and proteinwere studied. Total solids were determined by oven-drying. Single equations were unsatisfactory because of significant differences between the partial regressions for breeds. The most efficient equations included density and percent milk fat or protein within 6 milk fat intervals. For these, the standard errors of estimate ranged from 0.11-0.14% SNF, and fractions due to regression ranged from 0.60-0.77. The average deviations for annual herd averages among herds was 0.04% SNF, with 95% confidence limits of 0.10% SNF.