|Becerril C M, - CENTRO DE GANADERIA MEX|
|Wilcox C J, - UNIV OF FLORIDA|
|Sigmon K N, - UNIV OF FLORIDA|
Submitted to: Journal of Dairy Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: April 11, 1994
Publication Date: N/A
Interpretive Summary: In a subtropical environment, the percentage of white coat color in Holsteins affects production and reproduction. Evidence suggests that heritability of coat color is high when measured on a continuous scale, but no studies have been conducted on large field data sets from several herds using current statistical techniques that take advantage of modern computing capabilities. Although "restricted maximum likelihood" is the statistical method of choice today among animal breeders for estimating genetic parameters, this procedure requires data that are "normally" distributed, which is not the case for white coat color percentages. This study examined a data transformation procedure in an attempt to normalize the data and determine heritability of white coat color percentage. The trait was found to be highly heritable (>70%), which is higher than most economically important traits in dairy cattle. A transformation procedure that normalizes distribution of data is desirable when estimating genetic parameters for white coat color percentage. Genetic relationships of coat color with productive and reproductive traits may show that selection for increased percentage white may improve efficiency of dairy cattle in the subtropics.
Technical Abstract: Heritability of white coat color percentage was estimated from 4293 Holsteins on 8 Florida dairy farms. Estimates were by derivative-free restricted maximum likelihood using an animal model. Distribution of color percentages was not normal and was skewed to the right. Based on this study and on previous research, data were transformed by an extension of the Box-Cox transformation utilizing maximum likelihood. Heritability of white coat color percentage was .715 from untransformed data and .779 for transformed data. Standard errors of estimates were slightly lower (.0322 versus .0345) following transformation. Additional study to find an improved transformation still seems warranted.