I am also interseted in doing similar things. Talking from my
experience, I am not surprised that you are unsuccessful with mixing
two distributions (one binary-one normal or many binary- many normal
etc) in a bi-/multi-variate analyses. As far as I could recall from the
manual, ASREML is not equipped to do that.
I beleive that you wanted to acheive this with the model line "stay
!BIN !LOG pba1" .
The other model line you used "stay pba1 !BIN !LOG ~" is to fit bi-
variate threshold model (I do not think you intend to fit a threshold model
to a Normal trait as well). But let's say you had two binary traits and
you wanted to fit bi-variate-threshold model. Again as far as I could
recall you can't fit multiple-trait threshold models in ASREML (at least
not straight forward).
I am not very helpful here but like you, I would also be happy to
hear from ASREMLers about the above two problems (mixing
distributions and bi-varaite threshold models). Perhaps this is a
development that still need to be made within ASREML ?
> I am having difficulty successfully setting up a bivariate
> sire model with one (0,1) trait (stay) and one normally distributed
> trait (pba1).
> I suspect it is in the variance structures I am going wrong, but
> the ordering of the model line also makes a difference
> (ie stay !BIN !LOG pba1 ~......
> stay pba1 !BIN !LOG ~......)
> I get various problems - eg unable to invert R or G, likelihood diverging
> - depending on how I set up the R structure. It also doesn't seem to "see"
> the Tr.sire, only appearing to have one component for it when I can get it
> to run that far.
> The .as file follows. Any help greatly appreciated.
> Stayability to parity 3
> sire !P
> pbd1 pbdl
> avpba avpbd
> fgl 2
> pba1 pba2 pba3
> fg1 2 fg2 2 fg3 2
> pedig.mgs !MGS
> stay pba1 !BIN !LOG ~ -Trait Tr.fa !r Tr.sire !f Tr.fg1
> 1 2 1
> Tr 0 US 1 0 5.0
> Tr.sire 2
> Tr 0 US 0.1 0 0.7
> Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml