Hello!
I have a general question regarding the two-sample t-test. I have done a PPI analysis with 2 groups of subjects and wanted to test whether or not they were different and have entered the PPI contrasts from the first level into a 2 sample t-test. When I then estimate the model I either have two columns that are not independent from one another, or in one group nothing remains significant (when I look at that column alone), although when I analyse it separately in a one-sample t-test it is significant after correction from multiple comparisons (in the first case I left the variances equal, in the second case I made the variances unequal, but I don't understand why I should make the variances unequal). The two groups are different in size (the group that is smaller no longer shows a significant activation), and I have tried using both SPM5 and SPM8. From my understanding of statistics it should be possible to calculate the 2 sample t-test and the results should not be that different. I expect the groups to be the same but I wanted to test this statistically before continuing. Does anyone know why I need to have unequal variances in this case and why my results could be so different.
Thanks so much,
Virginia
|