Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)

First Advisor

Robert W. Adkinson


Over two million Holstein mating records was used to investigate selection and mating practices. Population statistics revealed non-random mating and selection. Herd production level was associated with higher breeding value sires and price paid per unit of semen. The decision to use an Artificial Insemination (AI) sire or natural service sire was studied and a discriminant function was derived that accurately categorized matings in a test data set of 117,000 observations, with an error rate less than 2.5 percent. Discriminating variables were herd production level, lactation number, service number, breeding month, cow status code, and days in milk at breeding. Natural service bulls were used to breed heifers and dry cows, but not milking cows---especially high producers. Selection of AI bulls was examined using stepwise regression on transformed frequency of bull use. Responses were regressed on 17 variables representing genetic and phenotypic characteristics of the bull. A model with 8 variables was selected using Mallow's coefficient. Variables included breeding values for production traits, final score, somatic cell score, and reliabilities. Major factors in AI sire selection were type, fat merit, net merit dollars, and somatic cell score. Mating of AI bulls and production cows showed that both tended to group independently of mate, based upon genetic values. Residual correlations reflected genetic correlations between production and type traits, except that there was a large negative association between type and production. This could have resulted from the way that production and type bulls have developed over time, or could indicate producers are willing to give up more production for type than was economically justifiable.