Premium
Phenotypic correlations as substitutes to genetic correlations in dairy sheep and goats *
Author(s) -
Kominakis A. P.
Publication year - 2003
Publication title -
journal of animal breeding and genetics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.689
H-Index - 51
eISSN - 1439-0388
pISSN - 0931-2668
DOI - 10.1046/j.1439-0388.2003.00397.x
Subject(s) - correlation , statistics , genetic correlation , sample size determination , biology , mathematics , standard error , genetic variation , zoology , veterinary medicine , genetics , medicine , geometry , gene
Summary A data set of n = 67 pairs of genetic and phenotypic correlation estimates for various milk traits, constructed from an exhaustive survey of published estimates in dairy sheep and goats, was used to test whether the phenotypic correlation ( r P ) may be a suitable replacement of the genetic correlation ( r G ). Various criteria to assess the closeness of the two estimates were used: Pearson‐, Spearman‐ and bootstrapped‐ correlations, paired sample t ‐test and non‐parametric tests. Correlations between paired genetic and phenotypic correlation estimates were generally found to be high (from 0.87 to 0.95) and not statistically different between the two species. Employment of a paired sample t ‐test showed that the average difference between the genetic and phenotypic correlation estimates (d = r G − r P ) was significantly (p < 0.05) different from zero only for dairy goats (d = 0.11). Comparisons of correlation estimates either after employment of Fisher's transformation or of a non‐parametric test (Kruskall–Wallis) showed that there was statistically no significant difference (p < 0.05) between the two estimates. The main sources of variation for pooled (sheep and goat) d were the year of publication, the method of estimation and the heritabilities of the traits. Furthermore, differences between the two estimates were caused by sampling errors of the genetic correlations. Although average reported standard errors of genetic correlations were found to be about 30% higher than the theoretically expected correlations, the difference between the two errors was not statistically significant (p < 0.05). A simulation study of two correlated traits displaying heritabilities, genetic, environmental and phenotypic correlations equal to the average literature estimates was also performed to investigate the effects of using r P instead of r G on the direct and indirect true selection responses and on the accuracy of the breeding values. Based on the results of the descriptive and simulation analysis for the species, traits and correlation structure examined here, phenotypic correlations may be substituted for genetic correlations when the latter are unavailable or are not precisely estimated.