Hi Veronika, Please, see attached how I'd code it, for use with randomise. Contrasts that are for within-subject effects (C1-C4, that test the effect of the condition within or across group) would need to include the option "-e design.grp", so that shufflings would happen within subject only. Contrasts that test between-subject effects (C5-C6, that test the interaction group vs. condition) would include not only the "-e design.grp", but also the option --permuteBlocks, so that blocks are shuffled as a whole. Ideally, there should be permutations within block together with whole block for an experiment as yours, and that will probably be available in a future version of randomise. All the best, Anderson On 18 July 2014 01:38, Veronika Vilgis <[log in to unmask]> wrote: > Hi, > > I would be very grateful for some feedback regarding my GLM model for a 3 > (group) x 2 (condition) Mixed Effect ANOVA. > I have been following the instructions based on the 2-groups, 2-levels per > subject (2-way Mixed Effect ANOVA) as described on the GLM section of the > fslwiki ( > http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/GLM#ANOVA:_2-groups.2C_2-levels_per_subject_.282-way_Mixed_Effect_ANOVA.29). > I also used this post to clarify the 3 group approach: ( > https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1311&L=FSL&P=R86271&1=FSL&9=A&J=on&d=No+Match%3BMatch%3BMatches&z=4 > ). > Currently the set-up is like this (only 2 subjects per group included): > > EV1(con) EV2 (conX1&2) EV3(conX1&3) EV4(conX2&3) > EV5(s1) EV6(s2) EV7(s3) ... > G1C1 1 1 1 > 0 1 0 0 > G1C2 -1 -1 -1 > 0 1 0 0 > G1C1 1 1 1 > 0 0 1 0 > G1C2 -1 -1 -1 > 0 0 1 0 > G2C1 1 -1 0 > 1 0 0 1 > G2C2 -1 1 0 > -1 0 0 1 > G2C1 1 -1 0 > 1 0 0 0 > G2C2 -1 1 0 > -1 0 0 0 > G3C1 1 0 1 > -1 0 0 0 > G3C2 -1 0 -1 > 1 0 0 0 > G3C1 1 0 1 > -1 0 0 0 > G3C2 -1 0 -1 > 1 0 0 0 > > > C1 1 0 0 0 > C2 0 1 0 0 > C3 0 0 1 0 > C4 0 0 0 1 > > I have a feeling that this approach isn't quite right. In the example on > the fslwiki group is ommited as an EV but 1 is assigned to group 1 and -1 > to group 2 which doesn't really allow for a third group. Is there a way > around this? > Any comments feedback would be greatly appreciated. > > Thanks! >