Dear Linling Li,
> I have two groups of data which all involves a same motor task. But the two
> groups of subjects have different frequencies in performing the task: one
> group has relatively small frequency. So if I want to compare the functional
> activation of the two groups, the first thing I need to do is to exclude the
> impact of different frequencies.
>
> At first, I did one-sample t-test for each group with the subjects'
> frequencies as a covariate. Then I did two-sample tests with a frequency
> vector which includes the frequencies of all the subjects (two groups) as a
> covariate for each group seperately. However, the group differencies seen by
> comparing the one-sample t-test results of two groups are really different
> from the results of two-sample t-test. Therefore, I'm wondering if a
> two-sample t-test with frequency as covariate is proper to exclude the
> impact of different frequencies. Or if there is some other way to exclude
> different frequencies' impact?
Chris is correct—if you have two effects that are highly correlated
(in this case, group and frequency), you won't be able to separate
them. Put another way, if group and frequency are correlated and both
in your model, then the two-sample t-test looking at the effect of
group will give you voxels where the independent effect of group is
significant—i.e., what can group explain that frequency can't?
If you have enough subjects, and any overlap in the frequency between
groups, you may be able to compare subgroups that are matched on
frequency. With or without the covariate, this would help disentangle
the two. With a reduced number of subjects it may be difficult to get
whole-brain corrected results, but you could also just do this
subgroup analysis in an ROI.
Good luck!
Jonathan
--
Dr. Jonathan Peelle
Department of Neurology
University of Pennsylvania
3 West Gates
3400 Spruce Street
Philadelphia, PA 19104
USA
http://jonathanpeelle.net/
|