Dear Michela,
> in a multisubject: conditions&covariates study for example, what
> are differences among:
>
> covariate:interaction? 1) none
> 2) with condition
> 3) with subject
>
Suppose you have a study with P subjects and Q conditions, and you
want to investigate the role of covariate A. If you select 'none', SPM will
create a single regressor, if you select 'with condition', it will create Q
regressors (one for each condition), and if 'with subject', P regressors.
Modelling (many) covariates reduces your d.f.s but may give a better fit of
your model, thereby decreasing residual error.
> WHICH one is the best and WHY?
> When I have to choose number (1) or (2) or
> (3) and WHY?
>
None is 'the best', that depends from your design and the question
you want to answer. As an example, if you have a multisubject PET-study with
8 scans per subject and you want to eliminate overall time (order) effects,
you could use 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 .... as a covariate
without cov x subj or cov x cond interactions. If your subjects have
performed a single task during each scan, and you want to investigate
correlations between individual performance and rCBF, you should specify
'interaction with subject'. If, on the other hand, in this example you have
both active (task) and baseline scans, you are interested primarily in cond
(task) x cov (performance) interactions, so choose 'with condition'.
> And what is the meaning of the covariate centre?
>
Basically, centering of a covariate is done so that only effects
within condition, subject or group are modelled and not part of the main
effect (for condition, subject or group). Generally it is suggested to
choose the default unless you have a reason not to (e.g., you're not
interested in main effects).
> Can anyone suggest me books or papers where I can find something about
> that?
>
To find more info, try the SPM-help facility (spm_spm_ui). There are
also some useful postings in the mailbase (e.g., Andrew Holmes' mail of
06-06-2000
(http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0006&L=spm&P=R3157).
Good luck
Dick Veltman
> --------------------------------------------
> Michela Canfora - Dipartimento di Fisica
> Via Dodecaneso 33 - 16146 Genova, Italy
> e-mail: [log in to unmask]
> Phone: +39 010 353 6314
> Fax: +39 010 311 066
> --------------------------------------------
|