Re: repeated measures models
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: repeated measures models

Dear Andrew,
Sorry to take so long to reply

The answer to youre query is not easy but here goes.
> hi, i'm having trouble working out how to fit repeated measures models
> in asreml. i have a data file with one record per "animal", with 3
> observations of the same trait, which looks something like this:
> tag intake1 intake2 intake3 ... (other fields).
> 1     x1      y1      z1    ...
> 2     x2      NA      z2    ...
> (i have missing values for some animals)
> i fit a model like this
> intake1 intake2 intake3 ~ Trait Tr.line Tr.feedflock Tr.damage,
>  Tr.brtype !r sire units
> 1 2 0 !ASUV
> 1728   #no. of animals in the data
> Tr

IF YOU ARE USING ASUV, your should explicitly include the mv factor
otherwise the missing values will be miss treated I think.

Check the degrees of freedom.  I would have expected this job not to work.

> this seems to work ok: i get variance estimates which look sensible.
> my next step is to add a second trait, which doesn't have repeated
> observations. i can't for the life of me think how to specify this
> model. what i would like to do is estimate "sire" and "residual"
> covariance matrices for "intake" and my second trait, and a permanent
> environment variance for "intake" (the "units" component from the above
> analysis). can anyone help me here?
> thanks
> -- 
> Andrew Swan
> CSIRO Division of Animal Production
> Pastoral Research Laboratory
> Armidale  2350  AUSTRALIA
> ph.  +61 (0)2 67761377
> fax  +61 (0)2 67761333
> email:

Lets start with 
  intake1 intake2 intake3 other ~ Trait Tr.line Tr.feedflock Tr.damage,
  Tr.brtype !r Tr.sire
 1 2 0 
 1728   #no. of animals in the data
 Tr 0 US  E11 E12 E22 E31 E32 E33 E41 E42 E43 E44  !=1212213330
 Tr.sire 2
 Tr 0 US  G11 G12 G22 G31 G32 G33 G41 G42 G43 G44  !=4545546660

SO what we have here is setting the model up as standard multivariate
and then constraining the intake variances and covariances.

You could of course fit it with unconstrained variances.

The way the first model was fitted, the 'units' term gave the E12 E32 R32
covariances and the  'units + residual' gave the E11, E22 and E33 terms.

The sire term gave G11=G21=G22=G31=G32=G33

When you get this running, you may see how to extend to more complex models.

Arthur Gilmour PhD                    email:
Senior Research Scientist (Biometrics)                 fax: <61> 2 6391 3899
NSW Agriculture                             telephone work: <61> 2 6391 3815
Orange Agricultural Institute                         home: <61> 2 6362 0046
Forest Rd, ORANGE, 2800, AUSTRALIA         

ASREML is currently free by anonymous ftp from pub/aar on
    Point your web browser at 
    in the IACR-Rothamsted information system 

To join the asreml discussion list, send the message  

The address for messages to the list is

                        <> <> <> <> <> <> <>
"Blessed are the pure in heart for they shall see God"  Jesus; Matthew 5:8