Dear Helmut:
Thanks for the reply. I agree with you that a paired t-test should be used here.
I am concerned that even I relaxed significant threshold of the paired t-test, the result was still different from that of two sample t-test. This worried me of the correctness of t-test. Therefore, I am seeking your opinions.
Thanks again!
Zuyao
________________________________________
From: H. Nebl [[log in to unmask]]
Sent: Friday, 27 March 2015 6:53 PM
To: [log in to unmask]; Zuyao Shan
Subject: Re: VBM: paired-test & two sample t-test
Dear Zuyao,
The contrast vectros are correct, but a two-sample t-test assumes two independent groups (thus also more df), which is not the case in your example. You have to go with a paired t-test because you have repeated measurements (reflected by the additional columns coding subjects, which results in less df).
Best,
Helmut
|