> X-Authentication-Warning: lamb.chiswick.anprod.csiro.au: petidomo set sender
to email@example.com using -f
> Date: Thu, 24 Feb 2000 11:26:20 +1100
> To: firstname.lastname@example.org
> From: Kim Bunter <email@example.com>
> Subject: SE's for boundary fixed parameters
> Mime-Version: 1.0
> Hi all,
> In several analyses fitting both additive and permanent environmental
> effects for animal (repeated records) ASREML has fixed one parameter
> (usually additive) to the boundary. I knew this might happen with my data
> as I do not have much pedigree depth or number of animals, but plenty of
> repeated records. For highly heritable traits fitting both effects was
> significantly better than either fitting one or the other, and the two
> parameters were successfully separated (according to LRT etc). The
> 'boundary' traits have the same pedigree and design structure as traits
> which converged normally, so there is obviously enough information in the
> data (at least for highly heritable traits), but I suspect that the
> 'boundary' traits are not heritable at all.
> So, my questions are:
> 1) does fixation of a particular parameter to the zero boundary indeed
> reflect that this trait is not heritable? Following on from this, can it be
> truly taken as evidence that the sole source of repeatability is the
> permanent environment of the dam? This is in light of the fact that I know
> estimates were obtained for both h2 and c2 for different traits under an
> identical design structure.
If the additive component hits the boundary, it usually implies that
the corresponding variance component is wanting to take a negative
value but the !GP qualifier is say that a negative value is 'out of bounds'.
It could be that there is another major source of variation in the data
which has been ignored, or that there is an interaction of the genetic effect
> 2) are the standard errors truly representative for the parameter which was
> sucessfully estimated? I know fixation can bias the remaining parameter
> estimates (although this was not the case for my data) and would assume the
> same might happen to SE's?
It is an asymptotic SE assuming the model is correct [ie assuming h2 is zero].
In the context you describe,
it sounds as if c2 and h2 might be fairly highly (negatively) correlated.
To test the sensitivity of the SE, why not fix the additive component
at a range of values and see how the SE of the other component changes.
> Thanks for your help
> Kim Bunter (M.Rur.Sc)
> PhD Student
> Animal Genetics and Breeding Unit
> University of New England
> Armidale, NSW, 2351
> Ph: (02) 6773 3788
> Fax: (02) 6773 3266
> email: firstname.lastname@example.org
> Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml
Arthur Gilmour PhD mailto:Arthur.Gilmour@agric.nsw.gov.au
Principal Research Scientist (Biometrics) fax: <61> 2 6391 3899
NSW Agriculture <61> 2 6391 3922
Orange Agricultural Institute telephone work: <61> 2 6391 3815
Forest Rd, ORANGE, 2800, AUSTRALIA home: <61> 2 6362 0046
ASREML is still free by anonymous ftp from pub/aar on ftp.res.bbsrc.ac.uk
or point your web browser at ftp://ftp.res.bbsrc.ac.uk/pub/aar/
To join the asreml discussion list, send the message
To send messages to the list, mailto:email@example.com.CSIRO.au
Asreml list archive: http://www.chiswick.anprod.csiro.au/lists/asreml
<> <> <> <> <> <> <>
"Christ Jesus came into the world to save sinners" I Timothy 1:15.
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml