 |
 |
Hi, You will have 14*4 inputs, but you will only have 17 EVs. You'll have a repeated measures ANOVA model. It is very important that all subjects have all 4 measures. If you have subjects with missing data for one of the 4 measures, they must be omitted (fancier model estimation techniques are needed for this and fMRI software can't handle it as far as I know). The trick is to start with the 2 way ANOVA model shown here http://www.fmrib.ox.ac.uk/fsl/feat5/detail.html#ANOVA2factors2levels BUT you need to account for the repeated measures, so omit their EV4 (column of 1s) and replace it with 14 subject-specific mean regressors, similar to what you'd use for a paired t test, but each subject has 4 measures instead of 2. If you forget to omit the column of 1s you'll get a "rank deficient" error message. The contrasts are the same as what is shown in the example at the link. Jeanette On Wed, Oct 13, 2010 at 6:02 PM, carlos silva pereira <[log in to unmask]>wrote: > Dear FSL list, > > I'm trying to implement an ANOVA, but I'm having difficulties in adapting > my design to the examples given in the FEAT manual. > > My experiment has 14 subjects that were scanned in 6 consecutive runs. The > same type and number of stimuli - 4 conditions plus a baseline, comprising > two factors each at two levels, were presented in each run. > > At a first level analysis, each of the six runs/participant was > individually analyzed. In the second level, the six runs of each participant > were fed into a fixed effects analysis. Several third level group t-tests > were then carried out for each desired contrast. > > My question is how to specify the design matrix, contrasts and F-tests, in > order to get the effect for the two factors and their interaction. > > What should be my inputs, 14*4 (14 subjects, 4 EV's)? > And should I have 4 + 14 EV's, in order to model each subject's mean? > > Also, if this is the case, how should I specify the contrasts and F-tests? > > Best regards, > Carlos >