| Author |
Message |
< ASReml ~ Correlations between estimated variance components
|
| lrn |
 Posted: Tue Aug 14, 2012 12:45 pm |
 |
|
|
Joined: 09 Aug 2012
Posts: 1
|
ASReml produces standard errors for the variance component estimates, but can I also retrieve the covariances between the estimates of the variance components? I want to check whether the estimates of two variance components might be highly correlated.
/Lars |
|
|
| Back to top |
 
|
| Peter Thomson |
| Posted: Wed Aug 15, 2012 2:25 am |
|
|
|
Joined: 20 Mar 2010
Posts: 4
|
Hi Lars,
I think you will find these values in the .VVP output file (in the standalone version)
Cheers, Peter |
|
|
| Back to top |
|
| bcullis |
| Posted: Wed Aug 15, 2012 3:46 am |
|
|
|
Joined: 22 Oct 2010
Posts: 45
|
Lars
If you are using ASReml-R these can also be found in the returned object list under $ai. This gives the matrix in lower triangle form, but there is a utility to produce the full (symmetric) variance matrix from this. It is given here below
Vmat <- asreml:::asreml.ltri2mat(obj$ai)
and if you wish to check this then you can simply do this
cbind(summary(obj)$varcomp[,'std.error'],sqrt(diag(Vmat)))
which should produce two identical columns. I hope there are no typos here & that this is helpful
cheers
brian |
|
|
| Back to top |
|
|
|
![[vsni.co.uk]](/common/images/address.gif){kind=small}
Contact us | Site map
News Feed
