To determine the extent to which heterosis estimates were confounded with each other or with other
fied effects, correlations among fied effect estimates
were computed by dividing the rows and columns of the
block of the inverse mixed model equations corresponding to those fied effects by the standard errors corresponding to those rows and columns, respectively as in
Thallman et al. (2014). Correlations with absolute value
close to one would indicate confounding. Similarly, to
identify colinearities among (co)variance components,
correlations among (co)variance component estimates
were computed by dividing the rows and columns of
the inverse approximate information matrix by the standard errors of (co)variance component estimates corresponding to those rows and columns, respectively.
Overall direct heterosis was not included in either model as the sum of the covariates accounting for
biological type heterozygosity are equivalent to the
overall direct heterosis. Contrasts among heterosis estimates of breed groups were obtained