Hi,
I have been running a bivariate animal model analysis
in which I have included a second genetic effect. This
genetic effect applies to both traits, but is highly
correlated.
I have done the analysis twice. In the first run I
estimated the unstructured covariance matrix without
any constraint.
ie the .as file contained the block:-
Tr.ss 2
Trait 0 US 0.13 0.07 0.08
ss 0 0
The correlation and its standard errorfor this effect,
from the .pvs file, were:-
1.1403 0.3020
In order to obtain an estimate within the parameter
space, I re-ran the analysis using the following
block in the .as file:-
Tr.ss 2
Trait 0 CORR 0.95 0.13 0.08 !GP
ss 0 0
In this case I get the following correlation estimate
and se from the .asr file:-
0.999 0.000
Why do I get a zero se for the correlation estimate
under the constrained CORRelation analysis? Is it
an artifact of having to keep that value less than
one, or is it indicative of there not being any
information for the correlation (in which case, why
wasn't there a zero se for the unstructured analysis
too?)?
Thanks in advance,
Ron.