Hi Aditya,
just to add to what Reza just said...
I think you are possibly confusing the uncorrected p-values (or equivalent t-values) used to threshold the initial t-maps to do the cluster-based correction for multiple comparisons (-c option) with the final, corrected p-values given by randomise.
If what you want is an initial threshold corresponding to a uncorrected p-value of 0.01, then you need to take a statistical table and see, with your dof, what is the corresponding t-value to put after the -c option. Using non-parametric testing, I was told that *any* threshold is valid (as opposed to parametric testing where a minimum of 2.3 is recommended), but your results might then be difficult to interpret.
As Reza said, the tfce option doesn't require this kind of thresholding consideration. Note that you can run both options at the same time...
Cheers,
Gwenaelle
> De: Aditya Kumar Kasinadhuni <[log in to unmask]>
> Objet: [FSL] Thresholding in VBM
> À: [log in to unmask]
> Date: Vendredi 1 avril 2011, 16h32
> Hello all,
>
> I was wondering, if we changed the cluster based
> threshold in randomize to a value greater than 2.3 (t
> value), say 3 then would the results of specifiying a
> threshold of 0.95 using fslmaths vary? I mean should we be
> applying a more stringent threshold (like 0.99) considering
> we used a greater value like 3?
> This might be a very stupid question but its kind of
> bothering me because of my results.
>
> Thank you.
>
> P.S: Just to be more clear, I am trying to use a threshold
> of p<0.01 for my images but have no idea of how to set
> this up in either cluster based or TFCE techniques. Is it
> the value -c 2.3 used in randomize that needs to be varied
> to -c 3.0 or the value in fslmath (-thr 0.95 to -thr 0.99)
> or both??
--------------------------------------------------------------------
Gwenaëlle Douaud, PhD
FMRIB Centre, University of Oxford
John Radcliffe Hospital, Headington OX3 9DU Oxford UK
Tel: +44 (0) 1865 222 523 Fax: +44 (0) 1865 222 717
www.fmrib.ox.ac.uk/~douaud
--------------------------------------------------------------------
|