The results should not be identical, if the covariate explains extra
variance in your data. SPM will fit the entire design matrix at once,
therefore there is no order effect. However, if your covariate explains
something that is already explained by (a linear combination of) other
regressors, then the change may be less than small.
Depending on which set of updates you use, there may be problems with
SPM not using the correct SPM.mat file. Make sure you have the latest
set of updates installed, use a different results directory for each of
the analyses and check that the files (SPM.mat, beta*img, con*img,
spmT*img etc.) are created in the right location.
Volkmar
Am Montag, den 04.08.2008, 17:14 -0400 schrieb D Chaws:
> Someone please tell me there is some info on this behavior. It would
> seem to be a major bug in SPM5 if flexible factorial designs did not
> account for the covariates entered in them. Just to recap on the
> problem, here is the issue:
>
> 1. Use a flexible factorial design
> 2. Define subject/main effects/interaction factors etc...
> 3. Run analysis and get result A
> 4. Include a critical covariate that should change the results
> 5. Re-run analysis and get result B [using same contrast matrix in A]
> 6. Result A is exactly identical to B, so the covariate has not been
> taken
> into account in the modeling procedure.
>
> Sorry if there is an obvious answer to this, or if I am missing
> something (completely possible). As it is, my only thought is that
> the covariates always seem to have to come last in the design matrix
> of these flexible factorial designs. Consequently, if sequential SS
> are used, I could see why the main effect of factor 1 that appears
> before the covariates in the design matrix is unaffected. Thanks in
> advance for your help.
>
> -- DC
>
> On Mon, Aug 4, 2008 at 6:12 AM, Greimel, Ellen <[log in to unmask]>
> wrote:
> No, sorry, I didnĀ“t a reply from the list,
>
> Dipl.-Psych. Ellen Greimel
>
> Child Neuropsychology Section
> Department of Child and Adolescent Psychiatry
> University Hospital Aachen
> Neuenhofer Weg 21, D-52074 Aachen
> Phone: +49-(0)241-80-89892
>
> Ph: +49-(0) 241-80-89892
> Fax: +49-(0) 241-80-82544
>
> Cognitive Neurology
> Department of Medicine, Institute of Neuroscience and
> Biophysics
> Research Centre Juelich
> Leo-Brandt-Str. 5
> 52425 Juelich
> Germany
>
> Ph: +49-(0) 2461-61-5820
> ________________________________________
> Von: D. Chaws [[log in to unmask]]
> Gesendet: Freitag, 1. August 2008 19:01
> An: [log in to unmask]; Greimel, Ellen
> Betreff: Re: full flexible anova spm5: covariate
>
> Did anyone ever figure this out? I just came across the same
> problem. In
> essence, the issue is:
>
> 1. Use a flexible factorial design
> 2. Define subject/main effects/interaction factors etc...
> 3. Run analysis and get result A
> 4. Include a critical covariate that should change the results
> 5. Re-run analysis and get result B [using same contrast
> matrix in A]
> 6. Result A is exactly identical to B, so the covariate has
> not been taken
> into account in the modeling procedure.
>
> - DC
>
>
--
Volkmar Glauche
-
Department of Neurology [log in to unmask]
Universitaetsklinikum Freiburg Phone 49(0)761-270-5331
Breisacher Str. 64 Fax 49(0)761-270-5416
79106 Freiburg http://fbi.uniklinik-freiburg.de/
|