Dear Nick,
We are comparing model 1 with LogL -330.82
Location 5 4.82
Loc.Cult 27 2.45
Loc.Cult.Rep 81 .35
Residual 2.16
With model 2 with LogL -315.13
Location 5 3.74
Cult 9 2.78
Loc.Cult 27 .00
Loc.Cult.Rep 81 .35
Residual 2.11
which fits significantly better.
The first model is fitting cultivars within locations as if they
were independent.
I.e. the Loc.Cult variance matrix (of order 27) looks like
2.45
.00 2.45
.00 .00 2.45
.00 .00 .00 2.45
.00 .00 .00 .00 2.45
.00 .00 .00 .00 .00 2.45
.00 .00 .00 .00 .00 .00 2.45
.00 .00 .00 .00 .00 .00 .00 2.45
.00 .00 .00 .00 .00 .00 .00 .00 2.45
etc
The second as if they are the same cultivar.
So the Cult + Loc.Cult variance matrix is
2.78
2.78 2.78
2.78 2.78 2.78
.00 .00 .00 2.78
.00 .00 .00 2.78 2.78
.00 .00 .00 2.78 2.78 2.78
.00 .00 .00 .00 .00 .00 2.78
.00 .00 .00 .00 .00 .00 2.78 2.78
.00 .00 .00 .00 .00 .00 2.78 2.78 2.78
etc
The model analysis is saying the latter is a better fit -
the cultivar effects are consistent across locations.
I suspect this is a case of applying what is true for fixed effects models
to random effects models and finding it does not hold.
I trust this helps
Arthur
> X-Authentication-Warning: lamb.chiswick.anprod.csiro.au: petidomo set sender
to asreml-owner@ram.chiswick.anprod.csiro.au using -f
> Date: Fri, 18 Dec 1998 15:33:02 +0800
> Mime-Version: 1.0
> To: asreml@ram.chiswick.anprod.csiro.au
> From: "N.W. Galwey" <ngalwey@cyllene.uwa.edu.au>
> Subject: Addition of term expected to make no difference
>
> I have run the .as file embedded below
>
> Graham Walton's canola data - oil content and yield
> Location 5 !A
> Soworder 4
> Cultivar 9 !A
> Rep 3
> Yield
> Oilcont
> Daysflr
> Flrtomat
> Temp
> Rainfall
>
> C:\DOCS\Si_P\GW\GWtotal.dat !skip 1 !maxit 20 !spline 10
>
> Oilcont ~ mu Rainfall lin(Soworder) !r dev(Soworder),
> Location Location.Cultivar Location.Cultivar.Rep
>
> (The corresponding .dat file is attached.)
>
> When I add the term Cultivar to the linear model, before the
> Location.Cultivar term, I get a substantial increase in the maximum
> likelihood. But I would expect that the addition of Cultivar would make no
> difference to the fitted values, but simply make the model more
> overparametrized.
You are only adding 1 parameter - a variance.
>
> Can the addition of a marginal term (changing a nested model to a crossed
> model) change the maximum likelihood?
YES
>
> Best wishes,
>
> Nick
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
Arthur Gilmour PhD email: 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
to asreml-request@chiswick.anprod.CSIRO.au
The address for messages to the list is asreml@chiswick.anprod.CSIRO.au
Asreml mailinglist archive: http://www.chiswick.anprod.csiro.au/lists/asreml
<> <> <> <> <> <> <>
"Do not be afraid, for behold,
I bring you good tidings of great joy which will be to all people.
For there is born to you this day in the city of David a Saviour,
who is Christ the Lord.
You will find the Babe wrapped in swaddling cloths,
lying in a manger." Luke 2: 10-12.
but the gift of God is eternal life in Christ Jesus our Lord." Romans 6:23
<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>