Skip to main content
ARS Home » Plains Area » Miles City, Montana » Livestock and Range Research Laboratory » Research » Publications at this Location » Publication #90662
Title: EFFECTS OF DIFFERENT STATISTICAL MODELS ON PREDICTION OF EXPECTED PROGENY DIFFERENCES

Author
item Macneil, Michael
item FERREIRA, G - UNIVERSIDADE FED DE SANTA
item GENGLER, N - BELGIUM
item Van Vleck, Lloyd

Submitted to: Governors Conference on the State of the Livestock Industry
Publication Type: Other
Publication Acceptance Date: 8/1/1997
Publication Date: N/A
Citation: N/A

Interpretive Summary: Recent advances in computing and statistical methods open many new ways to predict EPD. Now the question is how to best evaluate candidates for selection. Today, genetic evaluation relies on statistical models that account for the heritable part of genetic differences among animals. However, mating systems may also produce variation among animals that is not accounted for. No major changes of heritability estimates were observed when inbreeding effects were left out of models to predict EPD. Rankings of individual animals or sires also were not affected. Heritability estimates and the ranking of animals based on simpler versions of the animal model differed from results with the full animal model. The full animal model better separated direct and maternal genetic effects from each other and from non-genetic effects. Estimates of heritability were greatest from the full animal model. Lower estimates came from the parent models that included only sires or sires and maternal grand-sires. Correlations of EPD from sire models and the full animal model were also low. Thus, reranking of the animals evaluated was also important. Dominance effects may also exist for birth weight and weaning weight of Hereford cattle. Thus, when dominance effects make up part of the genetic variation, it may be important to fully include them in models used to predict EPD. These finds are interpreted to suggest that statistical models used for genetic evaluation be custom-tailored to the specific population being evaluated.

Technical Abstract: Recent advances in computing and statistical methods open many new ways to predict EPD. Now the question is how to best evaluate candidates for selection. Today, genetic evaluation relies on statistical models that account for the heritable part of genetic differences among animals. However, mating systems may also produce variation among animals that is not accounted for. No major changes of heritability estimates were observed when inbreeding effects were left out of models to predict EPD. Rankings of individual animals or sires also were not affected. Heritability estimates and the ranking of animals based on simpler versions of the animal model differed from results with the full animal model. The full animal model better separated direct and maternal genetic effects from each other and from non-genetic effects. Estimates of heritability were greatest from the full animal model. Lower estimates came from the parent models that included only sires or sires and maternal grand-sires. Correlations of EPD from sire models and the full animal model were also low. Thus, reranking of the animals evaluated was also important. Dominance effects may also exist for birth weight and weaning weight of Hereford cattle. Thus, when dominance effects make up part of the genetic variation, it may be important to fully include them in models used to predict EPD. These finds are interpreted to suggest that statistical models used for genetic evaluation be custom-tailored to the specific population being evaluated.