Cyril Pernet wrote:
> Hi chin wei
>
>>hi. I have some confusion about T and F statistic. Could you tell when
>>I use T statistic and when I use F statistic?
>>thanks in advance
>>
> There are no major computational differences between the F and the T.
> The difference is when looking to the distribution; the F test will look
> at differences for each side (two-tail) whereas the T test will look at
> one side (one-tail). It means for you that if you use a F test between
> two conditions C1 & C2, displayed results will be for C1>C2 and C1<C2.
> If you want to look at C1>C2, then you have to use a T test with the
> contrast [1 -1]. You can check in SPM: do an F test [1 -1] or [-1 1]
> it's the same ... and the two T tests [1 -1] and [-1 1] give the result
> of the F.
>
>
> Best,
> Cyril
In "Unified univariate and multivariate random field theory", K. Worsley
speaks about the difference between the two tests. One can see the
t-test as a test with one contrast whereas the F test is a test with
more than one contrast.
Matthieu
|