[log in to unmask] wrote:
> Dear Dr. Penny,
>
> may I trouble you with the following question:
>
> we performed a study in which both between- and within- subject factors
> have to be considered: there were two different groups (G1, G2) with each
> subject performing 6 different tasks (T1-T6)(within-subject factor).
>
> searching the archives, a previously suggested model for dealing with
> such mixed-effects was to set up a one-way ANOVA (without a constant term),
> specifying e.g. 12 groups (G1T1, G1T2, G1T3, G1T4, G1T5, G1T6, G2T1, G2T2,
> G2T3, G2T4, G2T5, G2T6).
>
> however, as our two groups differ in size (G1=12 subjects,G2=6 subjects)
> the ANOVA-model seems not to be applicable in this case.
>
I don't think this will be a problem. The part of the design matrix related
to the first group will have 12 rows and the part for the second group will have 6.
SPM will be able to handle this.
The only difficulty I foresee will occur if you use the nonsphericity
options. SPM may, naively, try to estimate covariances between all 12 'groups'.
If you do run into numerical problems you can switch off these
estimations (eg. covariances between the real groups should be zero) by specifying
no covariances between any of the 12 'groups' from the user interface.
Or, more complicatedly, you can kill some of the covariance basis functions (see earlier SPM
mails) which will prevent estimation of the specific covariances.
Best,
Will.
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7475
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|