By default, FS uses unsigned (ie, two sided) tests. If you have an a priori hypothesis about the sign, then you can use a signed test. When you use a signed test, the p-value is that of the unsigned test divided by two, ie, it is more significant. When you specify a sign for the MC correction, it changes the p-values to be that of a signed test so they are more significant than the unsigned test. But to do this you have to have an *a priori* reason for doing so. Once you've looked at the results, it is no longer a priori.
On 11/27/2015 09:48 PM, Karl Liu wrote:
Dear experts:
In examining QDEC results of a contrast of interest, I noticed the results are mainly in cold color (blue), with only two small clusters in warm color (red). When I proceeded with Monte Carlo simulation, if I chose the abs as the sign option in simulation, then no clusters survived. In contrast, some clusters would remain, if I chose the neg sign. Because the uncorrected results were mainly in cold color, it seems reasonable to me to use negative only in the simulation. But I am wondering how the option of abs/neg/pos leads to the disparities and will you please elaborate how the simulation works based on abs/neg/pos.
Thank you in advance.
Karl
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer