Hi,
Could someone please explain to me how the critical F value is obtained in SPM5?
It doesn't seem to be affected by the error degrees of freedom in a repeated measures
design, but it is affected by the effect df. Is this correct?
It appears to get smaller with increase in effect degrees of freedom (with larger number
of the contrasts vectors). Is this right?
So, if I specify p = 0.001 uncorrected,
and run an F contrast
contrast -1 0 1 -1 0 1 ==> Fcrit = 11.69
contrast -1 0 1 0 0 0; 0 0 0 -1 0 1; ==> Fcrit = 7.55
contrast -1 0 1 0 0 0; 0 0 0 -1 0 1; -1 2 -1 -1 2 -1 ==> Fcrit = 5.98;
contrast -1 0 1 0 0 0; 0 0 0 -1 0 1; -1 2 -1 0 0 0; 0 0 0 -1 2 -1 ==> Fcrit = 5.13;
I have a copy of Kiebe, Glaser & Friston 2003 paper on this topic, but I can't really follow
it.
By the way, would it then be valid to enter all the possible contrasts of interest in a
single F contrast; and then use this as a mask for more specific ones?
Thanks heaps in advance,
Branka
Branka Milivojevic
PhD candidate
Department of Psychology
University of Auckland
Private Bag 92019
Auckland
New Zealand
Ph: (+649) 373 7599 ext. 88636
Fax: (+649) 3737450
|