I just figured out something interesting that I thought would useful to share. I tried running a 2x2 repeated measures ANOVA in SPM, using the the flexible factorial module (four measures/contrast images per subject). The design matrix consisted of two columns for factor 1 two columns for factor 2, 4 columns for the interaction term, and a set of columns for subjects. I then generated a contrast for the interaction 0 0 0 0 -1 1 1 -1 … , and compared this to a one sample t-test of the image sums using the specified contrast (-1 1 1 -1). The two T-maps produced are quite different from one another, and the reason appears to be that the flexible factorial as described above is not properly specified. The correct model includes two extra terms: a factor 1 by subjects interaction term, and a factor 2 by subjects interaction term. The additional interaction terms are necessary to properly account for the within variance and acquire the correct error term. When those additional interaction terms are added in, the results become identical to the one sample t-test of the pre-configured image sums.
The simple solution here is to create the interaction contrast in advance and then do a one sample t-test, but the drawback is that doing so prevents you from being able to control/adjust for for unequal variance and independence, and given the within nature of the measurements, there is likely to be a sphericity violation.
You can specify the correct model using the flexible factorial module, you just need to be sure to do so or you will be using the wrong error term.
The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.