I am using a family model to look at the variance components of a binary
trait (single or multi-stemmed) in forest trees. Using the !bin qualifier
I get the variance components, but their standard errors bear no
relationship to their significance as judged with the LRT - SE's of <1 can
have a LL diference of >100!.
Presumably I can use the variance components directly in heritability
Is their any way of running a bivariate binomial model? Using the !bin
qualifier in a bivariate model causes ASREML to crash.