Production evaluation techniques based on lactation curves.

Abstract

Test day data on milk yield for individual cows were generated using Monte Carlo simulation consisting of 1000 herds. Each herd contained 30 second lactation cows for each of 2 yr. Three scenarios were simulated with increasing effects of test day and seasonality. For each test day, several statistics were calculated for each cow: test day data on yield deviated from expectations, deviated test day data on yield transformed to account for correlation of consecutive test day data on yields, 305-d mature equivalent estimates, and changes in these values from the previous test day. A probability value for each herd was calculated for test of month of lactation effects using ANOVA models with and without cows in the model. No month of lactation effects were simulated. The distribution of generated probability values were tested for uniformity using a chi-square test. The distribution of probability values associated with the change in test day deviations were most nearly uniform, and results for these variables were similar when the cow effect was removed from the analysis model. The transformed variables also provided a fairly uniform set of probability values, although interpretation of these statistical tests was more difficult. Tests based on mature equivalent, 305-d records were oversensitive.