Hi,
In one contrast of my group-level analysis I want to confine the analysis (correction for multiple comparisons) within gray matter rather than the whole brain to enhance statistical power. I used the following two methods:
(1).
I have an average gray matter mask (GM) obtained from all subjects using FAST. In Post-stats in feat gui, I entered GM as "pre-threshold masking" and then did FLAME.
(2).
I performed "randomise" by using the filtered_func_data of the contrast as input and GM as mask (using cluster-based thresholding).
The results are the attached figure:
Yellow = thresh_zstat (using average GM as pre-threshold masking) (>0)
Green = randomise result (using average GM as mask) (>0.95)
Blue = zstat (>2.5)
You can see that there are overlap voxels (i) between stats (Blue) and thresh_zstat (Yellow) and (ii) between stats (Blue) and randomise (Green). However, thresh_zstat (Yellow) and randomise (Green) images are different.
I had expected that both methods should yield similar results because they used cluster-based strategies......Could anyone tell me why and which method shows the real activations?
Thanks in advance.
Mark
|