 |
 |
Dear all (but most likely, Dr. McLaren), I'm sorry to revisit this issue -- I know it comes up over and over, and the answer is always "you need to use flexible factorial, or download GLM Flex." I just have a couple of (hopefully) clarifying questions. I have a fully within-subject, 3x2x2, where the 3 levels of the first factor are waves (i.e., measurements at different ages). Based on the advice of this forum, I plan to use the flexible factorial procedure and produce contrasts for main effects, 2-way interactions with the time factor, and simple effects of time within each level of the other two factors. My understanding is that flexible factorial correctly partitions the variance by subject, and uses the correct number of degrees of freedom. I also understand that Henson and Penny provide a way to correctly partition the variance using a full factorial approach here ( http://www.fil.ion.ucl.ac.uk/~wpenny/publications/rik_anova.pdf). My question is whether anything has changed as of SPM 12 that would let one correctly model the group level using full factorial with simple contrasts from the first level (as opposed the the Henson and Penny approach). I've heard somewhere that one doesn't usually see substantive differences whether one uses full or flexible factorial, so I'm wondering if I've missed an important aspect of previous discussions. Thank you so much for any advice, ~John John Flournoy, MS PhD Student, Psychology, University of Oregon Developmental Social Neuroscience Lab <http://dsn.uoregon.edu/> & Personality and Social Dynamics Lab <http://pages.uoregon.edu/sanjay/>