Dear SPM experts
I am currently playing around with the mask threshold
("defaults.mask.thresh") of the "spm_defaults.m"-File.
I used a simple visual stimulation and looked at the first-level GLM,
which revealed puzzling effects I cannot explain. Could you please
explain me these effects?
When I analyze the data with the default threshold (0.8), I receive a
very strong occipital activation - as expected. The same happens when I
lower the threshold to 0.0 and use an explicit mask. However, when I
analyze the data with threshold 0.0 and without explicit mask, the
activation suddenly disappears. The only activation found is at the
border of the recorded EPI (see attachments).
If I would have used a multiple comparison correction, this would be
understandable due to the increased voxel space. However, these effects
occur when using p<0.001, uncorrected. As far as I know, the 1st-level
GLM is a *mass-univariate analysis*, which means that each voxel is
analyzed independently. Why then are there differences when altering the
mask threshold? Am I right that this effect does not affect the
second-level analysis because it uses the con-files which are roughly
the same across all 3 conditions?
Thank you very much for your help!
Best,
Tobias Hauser
--
Tobias U. Hauser, lic. phil.
Department of Child and Adolescent Psychiatry
University of Zürich
Neumünsterallee 9
CH-8032 Zürich
Switzerland
