[1 -1; 1 -1] is redundant.
The reason an F-contrast is different than a t-contrast because as you mentioned it is non-directional. i.e. it is testing for either A > B or B > A, and not one or the other.
Also, the degrees of freedom are different, and a random variable that follows an F distribution is also the square of a R.V. following a t distribution. You might find it useful to consult a mathematical statistics textbook.
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Angela J [[log in to unmask]]
Sent: Monday, December 09, 2013 8:54 PM
To: [log in to unmask]
Subject: [SPM] F-test question
I'm aware that the F-test is non-directional and tests for positive or negative differences between conditions. But the F-contrast [1 -1; 1 -1] should test where condition A is greater than condition B, shouldn't it? So it doesn't really test for negative differences...
Also, why is the F-contrast [1 -1; 1 -1] not giving the same result as the t-contrast [1 -1] in a first level analysis?
Thank you in advance,