Hi all,
We are currently conducting a VBM study looking at the effects of anorexia
nervosa. We have run an initial analysis which did not include any
proportional scaling (e.g. it was looking at absolute grey matter
differences) and we found nothing. However, we have read other papers which
find no absolute differences, but do find differences after proportional
scaling. We are considering doing proportional scaling now, but we wanted to,
a) ensure that we actually have a good reason for doing it, b) ensure that we
understand why it is implemented as it is. So, here are our questions
regarding these issues:
1) In a previous post (Item #008135) John Ashburner wrote that proportional
scaling should be used depending on whether, "...you want to sensitise your
tests to regions containing differing amounts of grey matter, or differing
amounts of grey matter relative to the overall volume of grey matter within
the brain..."
So, we take this to mean that proportional scaling will allow you to pick up
small differences in a region that are large when the overall differences are
taken into account. Is this correct? If so, then would it be sensible to use
proportional scaling if you believe that your regional effects might be very
subtle?
2) We have seen others use a grand mean scaling value of 100. Why 100?
Thank you in advance for your help.
Cheers,
Blake Richards
|