Dear Ioan
> I want to determine whether fixed effects and interactions are
> significant from asreml output. The df for the effects are given.
> I am not sure about what "error" df to use when comparing the
> F values to values in a Variance Ratio table.
The denominator (error) degrees of freedom is not easily
calculated in a mixed model, which is why ASREML does not
report it. The numerator df is printed just before the
F ratio.
The denominator df will not be greater than the residual DF
reported in the LogL lines but will usually be less.
Sometimes knowledge of the structure of the experiment will help.
For example, in an RCB analysis, with t treatments and r reps,
the error df is (t-1)(r-1)
but if analysed as a mixed model with random rep effects, the
reported error DF is t(r-1)
In a split plot analysis you test some terms against error A and
some against Error B. So in a mixed model version, you use the
DF associated with the
appropriate strata (ASREML does not work this out for you).
In this case it is easy to work out what the proper tests are.
In general it is not but you can probably take reasonable guesses
which will be closer than assuming infinity (equivalent to
just doing a chi-square test ignoring error df).
> If I use F-incr do I use the following term's df ?
NO
> But what about the appropriate "error" df for the final term and all terms if
> F-adj is used ?
In any case, at the moment the best I can offer is to write down
the analagous Fixed AnovA table and the df it would have, and use those.
This is not easy and in the unbalanced case will never be exact
but will usually be close enough to draw a valid conclusion.
Recent papers are Kenward and Roger 1997 Biometrics 53: 983-997
and Elston 1998 Biometrixs 1085-96
Arthur
> From: Dr Ioan Ap Dewi <i.apdewi@bangor.ac.uk>
> To: "'Arthur Gilmour'" <gilmoua@ornsun.agric.nsw.gov.au>
> Subject: testing fixed effects
> Date: Wed, 26 May 1999 09:20:49 +0100
> MIME-Version: 1.0
> Content-Transfer-Encoding: quoted-printable
> Content-Transfer-Encoding: quoted-printable
>
> Arthur
>
> I want to determine whether fixed effects and interactions are
> significant from asreml output. The df for the effects are given.
> I am not sure about what "error" df to use when comparing the
> F values to values in a Variance Ratio table.
>
> If I use F-incr do I use the following term's df ?
> But what about the appropriate "error" df for the final term and
> all terms if F-adj is used ?
>
> Thanks
>
>
> Ioan Ap Dewi
>
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Arthur Gilmour PhD mailto:Arthur.Gilmour@agric.nsw.gov.au
Senior 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
subscribe
mailto:asreml-request@chiswick.anprod.CSIRO.au
To send messages to the list, mailto:asreml@chiswick.anprod.CSIRO.au
Asreml mailinglist 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