Thanks very much for your reply. I think I did not explain my model clearly enough. In order to be clear, I used 2 individual in each group as example in my first post. The actual number of participants in each groups are: D1=71, D2=206, A1=213, A2=64. I used FEAT rather than randomise to get the significant F of mean diagnosis without any significant in both related t-test (C2 and C5). Do you think it would be useful if I run my model with randomise due to strongly unbalance the group?
Furthermore, such problem existed on another model where I tested A2 group mean by FEAT (F1 significant and none of t-test significant). This was the only group showed wired results when I tested group mean D1, D2, A1, A2. Do you have any suggestion on this please? Sorry about so many basic questions. Thanks a lot.
EV
1
1
...
1
1
contrast F1
1 X
-1
|