Multitrait Random Regression

Dear Weiguo,
On Wed, 2008-12-31 at 11:31 -0600, Weiguo Cai wrote:
Hello,
>
> My name is Weiguo Cai, a asreml user at Iowa State University. I
> would like to subscribe ASReml discussion list. I have a couple of
> questions when I run genetic analysis using multi-trait random
> regression on asreml.
* Hopefully Damian will attend to that but you should join the ASReml forum
at http://www.vsni.co.uk/forum where I will post this reply.

> I still struggle on a couple of questions on variance structure
> specification, although I have looked through manuals for several
> days. I have 4 traits (FI, WT, BF, and LMA), which are all
> longitudinal measurements. First, the manual says we could even fit
> more than 2 direct product variance model for either R or G
> structure. I fitted the random term as (!r leg(age,1).Trait.Anim) and
> specified the 3 direct product variance model for it, but it won't
> work(asreml failed to read G structure line). I am wondering if the
> asreml support the 3 direct product variance model or not? I also try
> to fit a heterogeneous residual variance on each month for one of
> traits, rather on each day of that measurement. I am also wondering
> if asreml could group the residual variance and each group have their
> own residual variance. I will highly appreciate it if you would give
> me a response.
>
> Best regards,
>
> Weiguo Cai
>
>
> ASReml can handle 3 way structures in some cases so the error you have is syntax (you have not specified it properly). However, it is unlikely that a 3way structure is what you want in this case. Rather, I would try something like

leg(age,1).Trait.Anim 2
0 0 US !GP
.
. .
. . .
. . . .
. . . . .
. . . . . .
. . . . . . .
. . . . . . . .

Anim 0 AINV

where you would supply the initial vales for the 8x8 genetic matrix for
4 intercepts and 4 slopes

Unless you have a large data set, it is highly probable that the estimated matrix will be difficult to estimate (i.e. be on the
boundary of the parameter space). You could start with values
from the four univariate analyses.

I would explore heterogeneity of residual variance in univariate models first. It would be tricky to set up the model with month specific variances
for just one trait in a multitrait analysis. The approach would be to
define the model allowing month specific error matrices for all traits and then constrain some to to be equal (across months).

Arthur

------------------------
Arthur Gilmour

Principal Research Scientist (Biometrics)
NSW Dept Primary Industries

-------------------- m2f --------------------

Sent using Mail2Forum (http://www.mail2forum.com).

Read this topic online here:
http://www.vsni.co.uk/forum/viewtopic.php?p=340#340

-------------------- m2f --------------------
Received on Sat Jan 02 2009 - 04:51:27 EST