All,
If I understand the subject modes correctly, a postive subject mode value is
given to a subject with positive regression coefficients that have been
summed and averaged over the whole brain compared to the rank-1 timecourse.
Or to rephrase, the factor loadings of the subjects' timecourses with
respect to the rank-1 approximation/largest eigen vector timecourse.
Negative subject modes represent the inverse of this. So, for resting data
it would make sense to take the absolute value of the subject modes, using
the concat approach (just an example, I don't propose doing this).
The colors on the main display output still confuse me though
because most spatial maps contain orange/red values and SELDOM blue ones,
even if the subject mode values are distributed evenly with respect to being
positive or negative as they often are with resting state data. Sometimes in
the list, blue values are referred to as being the result of
anti-correlations. Since subject modes can be positive or negative and the
spatial regions on the map can be positive or negative, I seem to be missing
something;) In brief, in seems that even though half of the subjects have
negative subject mode values, all/most of the information in the spatial
maps is regarding positive regression coefficients, I would think the
spatial maps would be more evenly split between red and blue values. Is the
absolute value taken somewhere, because otherwise the subjects with positive
and negative subject modes would tend to cancel each other out when it comes
to spatially significant results. Thanks for any help.
Perhaps the confusion is that the spatial maps are made using absolute
values of regression coefficients while the subject modes are the result of
factor loadings?
Chris Bell
|