 |
 |
Hi all, I've been playing with a 2x2x2 ANOVA with two within subject factors and one between subject factors, along with a separate factor for subject (since these are main effects at the first level) using a full-factorial design. I've since tried running a smaller 2x2 ANOVA with a single within subject factor and a single between-subjects factor, attempting to replicate the approach detailed in the Gittleman ANOVA primer. Instead, however, I get a really odd design matrix (attached), that doesn't look like what it should, nor look like a portion of the 2x2x2 ANOVA, which looked right. To get this design matrix, I entered the vector for the conditions much as I had for the 2x2x2, except with the last column missing (e.g for the two scans for the within-subject factor from a subject belonging to group 2, where the first factor I specified after the subject factor is the between-subject factor, I would enter [2 1;2 2]. If, however, I invert the columns and enter [2 2;1 2], I would get a correct- looking design matrix, though this seems to not be doing things correctly. In looking at the design matrix, the first group's subjects look OK, but it's the second group that do not when entering things the first way (which would make sense since for the first group I would enter [1 1;1 2] whether I were doing things the initial way I tried or the inverted matrix way. I also tried respecifying, starting from scratch, etc, but always came to the same result. Any thoughts? many thanks, Amit