 |
 |
Hi Max, It's a bit hard to see the headers are for each of the EVs in the first design but it seems to me that that is the one you want. That is, all between-subject factors are modelled by the subject-specific EVs (EV8 onwards in your design), whereas EVs 1 through 7 model differences between the various conditions, main and interactions. These two give different results because they model the responses in very different ways. The 2nd design doesn't take care of the between-subject variability. All the best, Anderson On Tue, 28 Aug 2018 at 20:41, Max Levinson <[log in to unmask]> wrote: > Hi all, > > I would like to set up a group-averaged 2x4 repeated measures ANOVA (with > FLAME1). Every subject has the same design, and I am interested in group > averages of the effects (main and interactions). I came up with two > different implementations of the final group-level design, and I'm hoping > for some input on their validity. In the example designs attached/below I > will pretend that I only have 2 subjects, for simplicity. > > The factors correspond to stimulus conditions in an event-related design; > say A1 A2 for first factor, and B1 B2 B3 B4 for the second factor. A > particular stimulus could be A1B1 or A1B2, for example. For each subject, > I've already modeled each stimulus condition. The outputs are 1 COPE per > condition (A1B2, e.g.), for a total of 8 COPES. The next step is to combine > across subjects in the 2x4 ANOVA design. > > I see two possible ways to do so, and I'm hoping for an explanation for > why they produce different results: > > 1. The first way, as suggested by some discussions in this forum, is to > input all COPES (8 per subject) and account for subject means in the design > (see method1.png attached) > Inputs: copes 1-8 for subject 1, copes 1-8 for subject 2 > > 2. The second way is to create all of these contrasts at the > individual-subject level first (same design as previous, but with only one > subject's inputs) and then use a single-group average design for the > top-level analysis. (see method2.png attached) > Final Inputs: 7 COPES per subject, corresponding to the 7 contrasts > created in the previous design. > > The two methods give pretty different effect sizes (before statistics), > and I'm not so sure why this is the case or what the implications are. I > would like to know whether the second method is valid or if there is a way > to adjust it to make it as valid as the first method. > > Thanks very much, > Max > > ######################################################################## > > To unsubscribe from the FSL list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 > ######################################################################## To unsubscribe from the FSL list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1