 |
 |
Dear SPM Community, I have several groups that I would like to compare. However I have some doubts regarding the contrasts. Just to make it simpler imagine 3 groups for a volumetric analysis, without any covariate. Let's say I want to compare group 1 with group 2 (eg. G1>G2) and then group 1 with group 3 (eg. G1>G3). Is it more "correct" to use, as a model, separate analysis with: - t-tests Contrasts: [1 -1] - same for groups 1 and 2 and then for groups 1 and 3 - or to perform an 1way-ANOVA (that includes all 3 groups in the design matrix) and then perform t-contrasts Contrasts: [1 -1 0] and [1 0 -1] For sure the results are different, since in the second analusis (ANOVA), the results for each contrast have much more degrees of freedom than in the first analysis (individual estimates of t-tests). I don't know if I was clear about my doubt. I would appreciate some advice/input regarding this issue. Thank you! Best regards, OtÃlia