Dear Stéphane, I'm afraid you're asking more than FSL can deliver: You have a design with a mix of with and between subjects factors. As Donald McLaren has expounded upon in many postings, FSL's approach to group modelling only allows a single measure per subject, *or*, if you have multiple measures per subject,you have only within subject effects (e.g. a 1-way repeated measures ANOVA... 1 group of subjects, each subject with k measurements). This design, with a mix of within- and between-subject factors can't be fit in Feat or Randomise correctly. Sorry! -Tom On Thu, Feb 9, 2012 at 9:43 AM, Stéphane Jacobs <[log in to unmask]>wrote: > Hello list, > > I have a 2x3 factorial design, with one factor with 2 levels, and the > other with 3 levels. As several persons before me, from what I've seen in > the archives, I've been having a hard time understanding how to set up my > analysis, but I think I've figured it out, and would love to have a > confirmation that what I do is correct before running it. > > So, I've used the 2x2 ANOVA and the tripled t-test examples to set up my > EVs. EV1 corresponds to factor A (2 levels), while EV2 and EV3 correspond > to factor B (3 levels). Then, I obtain EV4 and EV5 by multiplying EV1 and > EV2, and EV1 and EV3, respectively. If I'm correct, EV4 corresponds to the > interaction between factor A and the 2 first levels of factor B (AB[1-2]), > while EV5 corresponds to the interaction between factor A and level 1 and 3 > of factor B (AB[2-3]). The subsequent EVs then represent each individual > subject. The following table illustrates this for 2 subjects for the sake > of simplicity: > > > > EV1 EV2 > EV3 > EV4 > EV5 > EV6 > EV7 > A1B1 (s01) > -1 > 1 > 1 > -1 > -1 > 1 > 0 > A1B1 (s02) > -1 > 1 > 1 > -1 > -1 > 0 > 1 > A1B2 (s01) > -1 > -1 > 0 > 1 > 0 > 1 > 0 > A1B2 (s02) -1 > -1 > 0 > 1 > 0 > 0 > 1 > A1B3 (s01) -1 > 0 > -1 > 0 > 1 > 1 > 0 > A1B3 (s02) -1 > 0 > -1 > 0 > 1 > 0 > 1 > A2B1 (s01) 1 > 1 > 1 > 1 > 1 > 1 > 0 > A2B1 (s02) 1 > 1 > 1 > 1 > 1 > 0 > 1 > A2B2 (s01) 1 > -1 > 0 > -1 > 0 > 1 > 0 > A2B2 (s02) 1 > -1 > 0 > -1 > 0 > 0 > 1 > A2B3 (s01) 1 > 0 > -1 > 0 > -1 > 1 > 0 > A2B3 (s02) 1 > 0 > -1 > 0 > -1 > 0 > 1 > > Then, the contrasts and corresponding F tests to get the classical omnibus > test would be as follows: > > > EV1 > EV2 > EV3 > EV4 > EV5 > EV6 > EV7 > > F1 > F2 > F3 > A > 1 > 0 > 0 > 0 > 0 > 0 > 0 > > X > > > B1-B2 > 0 > 2 > 1 > 0 > 0 > 0 > 0 > > > X > > B1-B3 > 0 > 1 > 2 > 0 > 0 > 0 > 0 > > > X > > B2-B3 > 0 > -1 > 1 > 0 > 0 > 0 > 0 > > > X > > AB[1-2] > 0 > 0 > 0 > 1 > 0 > 0 > 0 > > > > X > AB1-3] > 0 > 0 > 0 > 0 > 1 > 0 > 0 > > > > X > > > Thanks for the help and great support, as always! > > Stéphane > > -- > Stéphane Jacobs - Chercheur post-doctorant / Post-doctoral researcher > > ImpAct - Inserm U1028 - Equipe Pélisson > Centre de Recherche en Neurosciences de Lyon > 16 avenue du Doyen Lépine > 69676 Bron Cedex, France > Téléphone / Phone: (+33) (0)4-72-91-34-20 > > -- __________________________________________________________ Thomas Nichols, PhD Principal Research Fellow, Head of Neuroimaging Statistics Department of Statistics & Warwick Manufacturing Group University of Warwick, Coventry CV4 7AL, United Kingdom Web: http://go.warwick.ac.uk/tenichols Email: [log in to unmask] Phone, Stats: +44 24761 51086, WMG: +44 24761 50752 Fax: +44 24 7652 4532