Hello!
I am trying to use the z-map generated by FSL IRVA analysis with a clustering
program of our own. It seems that F values that would generate negative z scores
are not tranformed (or are masked to zero). I tried to transform the F values to
z using the cephes library (used by FSL as well) functions ndtri(fdtr(DoF1,
DoF2, F)). It seems to produce inconsistent values compared to ftoz in FSL. We
have 3 repetitions with 30 images in event. The degrees of freedom should
therefore be 30-1 and 30*(3-1)?
The maximum value in FSL F-map is 5.17292 and in z-map 4.22877.
ndtri(fdtr(29, 60, 5.17292)) gives 5.36788717.
I would appreciate if someone could clarify the F to z transformation and why z
map is masked. Am I using wrong degrees of freedom. What about the filtering
effects on degrees of freedom?
-Antti Korvenoja
----------------------------------------------------------------------------
Antti Korvenoja
Department of Radiology & BioMag Laboratory
Helsinki University Central Hospital P.O. Box 700, FIN-00029 HUS, Finland
tel:+358-9-471 71786 * gsm:+358-40-7498469 * fax:+358-9-471 75781
[log in to unmask] * http://www.helsinki.fi/~akorveno/
----------------------------------------------------------------------------
|