 |
 |
Dear experts, I have run several statistical tests and was surprised to see unexpected differences in their results. Specifically I ran three types of tests: two way ANOVA using the 'full factorial' function, one sample t and paired t test. I wanted to compare brain activity in task A (modulated by regressors using parametric modulation) and task B (modulated by regressors using parametric modulation) in drug vs. placebo conditions (all within subject design) and thus I first ran two way within ANOVA. Specifically, 1) I chose the "Full factorial" 2) specified two factors (1. condition (2 levels: drug and placebo), 2. task (2 levels: task A and task B)) 3) I put "No" under "Independence" for both factors as my study is a within-subjects design 4) everything else was left as it was 5) Cells: [1 1] - I put contrast images of task A in drug condition, [1 2] - contrast images of task B in drug condition, [2 1] - contrast images of task A in placebo, [2 2] - contrast images of task B in placebo Now, in addition to the T and F contrast images created automatically by SPM, I defined a new T contrast to see brain areas of activation in task A vs. task B (both tasks were modulated by regressors) in drug condition by putting [1 -1 0 0]. In order to be sure I ran two additional tests: 1) I ran one sample t test using [task A - task B] contrast images in drug condition. 2) I ran a paired t test using contrast images from each condition (i.e., task A contrast images and task B contrast images of drug condition). I have found the the results from the two tests are similar (although not identical) but they are dramatically different from the results I got from the T contrast I created in ANOVA. Did I do something wrong? How could they be so different? Any insight would be greatly appreciated!, Claire