> > what kind of <brain_mask> and <cluster_prob_thresh> should be?
>
>a typical <cluster_z_thresh> to use is 2.3 and <cluster_prob_thresh> is
>your false positive rate (e.g. 0.05), but refer to the standard Gaussian
>Random Field Theory literature/discussions on the issues behind choosing
>these values.
Can I just add that choosing an appropriate cluster_z_thresh and
cluster_prob_thresh is a non-trivial issue and depends entirely on what
question you're asking and what type of activation you expect. This was
discussed in some detail a few months ago on this list. You might look
through the archives for a more complete discussion.
Briefly, for exploratory purposes a corrected p value of 0.05 is *very*
conservative. If you are simply exploring your data, I'd suggest a low
height threshold such as Mark's suggestion of 2.3 and a high p-value (like
1.0). For reporting data in most applications, a height threshold of 2.3
is far too low. It will produce enormous, anatomically implausible
clusters. For that you'll need a higher z-value (Z>3.1 or higher) and a
reasonable p-value (such as p=0.05).
Good luck,
Joe
|