I have a question as to how the small volume correction works in SPM12. If I run an uncorrected analysis with a p threshold of .001 and add a mask for a specific region, I believe this then calculates significant voxels using data pertaining to the whole image, but only shows me significant results within my region of interest. This is confirmed by my volume dropping form e.g. 446896 voxels to 37746 voxels. If I then click on "small volume" I can limit the number of comparisons to this masked region, making any corrections less conservative. However as I have not applied a correction for multiple comparisons, I am struggling to understand how this small volume correction is appropriate. I can see it changes the FWE and FDR corrected p values in the output, which makes sense as they are calculated dependent upon the estimated number of independent observations, which is reduced with a small volume correction.
Critically I wanted to know, when I apply a small volume correction does it automatically apply some non FWE or FDR correction for multiple comparisons (based on Random Field theory perhaps?), meaning that my originally uncorrected output is now meaningfully corrected, or is it always to be interpreted alongside another correction for multiple comparisons such as FWE?
|