Re: multivariate genetic analysis using SPLUS asreml

Dear Jarrod and Alastair,
Thanks for the advice. Changing ~1 to ~trait doesn’t seem
to do anything but changing the G structure for Dam and Cage
does change the estimates. Unfortunately, it does not
eliminate the scale problem. Calculating the heritabilities
using a single trait model gives consistent answers (no
scale effect) that are essentially the same as those from a
standard anova analysis. I know that there is nothing
fundamentally wrong with the data as a manova analysis gives
sensible results. It is not clear if I am making an error
in my coding or there is a problem with the program itself.
Unhappily the splus version seems to be poorly supported and
I am on the verge of giving up on it.
Thanks for all your help.
Derek

---- Original message ----
>Date: Wed, 17 Oct 2007 18:44:15 +0100
>From: Jarrod Hadfield <J.Hadfield_at_ED.AC.UK>
>Subject: Re: multivariate genetic analysis using SPLUS
asreml
>To: ASREML-L_at_AGRIC.NSW.GOV.AU
>
> HI,Â
> The main problem is perhaps the single intercept for
> the bivariate response (replace "~1" with
> "trait"). I think this should solve the problem,
> although it may still be necessary to rescale if the
> traits have v.high or v.low variances on their
> natural scale.
> Also, the appropriate G structure for Dam and Cage
> are probably multivariate (replace Dam with
> us(trait):Dam):
> HS.asr <- asreml(cbind(TRAIT1,TRAIT2)~trait,
> Â Â Â Â Â Â Â Â Â
> random=~us(trait):ped(Animal)+ us(trait):Dam +
> us(trait):Cage,
> Â Â Â Â Â Â Â Â Â rcov=~units:us(trait),
> Â Â Â Â Â Â Â Â Â
> ginverse=list(Animal=AINV), data=Data)
> Jarrod
> On 17 Oct 2007, at 18:14, Derek Roff wrote:
>
> HS.asr <- asreml(cbind(TRAIT1,TRAIT2)~1,
>
> Â Â Â Â Â Â Â Â Â
> random=~us(trait):ped(Animal)+ Dam + Cage,
>
> Â Â Â Â Â Â Â Â Â rcov=~units:us(trait),
>
> Â Â Â Â Â Â Â Â Â
> ginverse=list(Animal=AINV), data=Data)
Derek Roff,
Department of Biology,
University of California,
Riverside,
CA 92521
USA
Received on Sun Oct 17 2007 - 14:15:05 EST