Repository logo

Browsing by Author "Bourdon, Richard M., advisor"

Now showing 1 - 1 of 1
  • Thumbnail Image
    ItemOpen Access
    Heifer pregnancy genetic prediction and simulation modeling techniques
    (Colorado State University. Libraries, 2008) Comstock, Carlton R., author; Enns, R. Mark, advisor; Bourdon, Richard M., advisor
    The Colorado Beef Cattle Production Model (CBCPM) was rewritten with an object oriented design and used to simulate heifer pregnancy data with varying levels of age at puberty (AAP), probability of conception (PCON), and length of breeding season. Five percent of the heifers were simulated infertile due to non-genetic causes. Simulated data were used to estimate heritability of heifer pregnancy and to obtain EBV using threshold models. The EBV were tested for accuracy of prediction of the simulated genetic fertility traits. Object oriented methods used illustrated the ability of these techniques and tools, such as Unified Modeling Language, at organizing complex processes in ways to reduce errors and code maintenance effort, and to facilitate collaboration among developers. Adoption of these tools will be critical to the advancement of systems models. Heritability of heifer pregnancy from 20,000 heifers with very early puberty in the first 25 d of the breeding season was .139, .107, and .143 for mean PCON of 60, 70, and 80%, respectively, close to the .10 input heritability of PCON. The higher estimates may be due to a few heifers having two opportunities to breed. With very late puberty and 80% mean PCON in a 25 d breeding season the heritability estimate of AAP was .337, lower than the simulated .40 heritability. The estimate was lower because there was not 100% conception, some heifers were infertile, and puberty was observed as a threshold trait. Heritability estimates of heifer pregnancy generally decreased as breeding season length increased, likely due to an increasing percent of open infertile heifers. The ability of the variance component estimation software to converge on an estimate decreased as the number of open, fertile heifers decreased. The most difficulty was with high PCON, early puberty, and long breeding seasons; only one out of 100 estimates converged at 340 d AAP, 80% PCON, and 120 d breeding season. Calculated accuracy for heifer pregnancy EBV for the sires of the heifers using prediction error variances from a linear model, with the binary pregnancy observations treated as continuous data, overestimated accuracy of the EBV with respect to the simulated traits in all cases. Calculated accuracy was insensitive to changes in frequency of heifer pregnancy observations. Accuracy calculated as the simple correlation of the EBV with each simulated fertility trait for the sires of the heifers was highest in most cases at the shortest breeding season. The correlation with AAP was essentially zero for early puberty, and strongest (-.775) at late puberty. As breeding season length increased the AAP correlation declined toward zero. The accuracy for PCON was less sensitive to changes in AAP and breeding season length, ranging from .146 to .753; the strongest correlations were with early puberty and low PCON.