Bias in composite milk samples with unequal milking intervals.

Milk samples from a herd with milk intervals approximately 14-h nights and 10-h days were used to examine potential bias in measurements with composite samples. Six consecutive milkings were weighted and sampled individually. Composite samples for a day were determined mathematically from the data. Composite samples with equal amounts from a.m. and p.m. milkings gave fat percentage measurements that were .07% above the true daily measurement. Using the milking interval to determine the portion of sample from each milking reduced this bias to -.02%. Protein percentage and SCC were satisfactory with equal sampling (mean bias of .01% and 7, respectively) but were improved slightly with milking interval sampling regimen (mean bias of .00% and 1, respectively). Fat yields were determined with equal portions, milking interval portions, and two plans available from DHIA, namely, alternate component sampling and alternate a.m.-p.m. Based on mean deviation from true fat yield, alternate a.m.-p.m. was as accurate as alternate component sampling, but variations were greater with alternate a.m.-p.m.