Development of Predictive Equations for Milk Yield and Dry Matter Intake in Lactating Cows

Eleven cooperating university agricultural experiment stations provided information on 19 separate experiments in which total feed intake and composition, body weight, season, stage of lactation, milk production, and milk composition were recorded. A total of 492 cows on these experiments provided 4135 Holstein and 704 Jersey 28-day records. By stepwise backward elimination procedures, 192 separate regression models were studied to identify important independent and dependent variables in prediction models for milk yield and dry matter intake. Then by least squares procedures predictive models were developed for milk yield (squared multiple correlation .807) and dry matter intake (.741). A second independent experiment of 480 observations on 40 cows tested the validity of these predictive models. Correlations between predicted and actual values were .76 and .75 for milk yield and dry matter. The percent error in comparing estimated with actual observations was 4.9% for milk yield and 12.5% for dry matter intake. These errors could be significantly reduced by minor adjustments of models.