Hello Josselin,
I'd also have difficulties interpreting this specific contrast, I must say.
- First, I'd rather use explicit total GM as correction variable (potentially,
ICV in addition), if you are interested in regional GM effects. I think
this has
been discussed a lot before -- for the interpretation I think a "simple"
multiple
regression model is easier as you do not have to think about how the prop
scaling turns the relationship between you variables of interest and GM...
- You used proportional scaling instead -- would be interesting
to see, what JUST the difference is between prop scaling and
total GM as covariat (with all other variables left unchanged).
- So, usually, the general GM decline of regions is contained
in the global GM regressors (setting a positive contrast on total GM
=> high positive T-values)
- Indeed, then asking for a + contrast on age rather gives you areas that
produce a positive slope AFTER correction for its change correlated with total
GM
- We have seen strong relative preservation of (modulated) GM and indeed
interpreted them as "slower than average aging", or in the case of the neg.
contrast, "faster than average aging". (in absolute terms, so, if not
corrected
for total GM [or prop scaling], you'd see that in these areas there is also
decline
of volume with age...)
- To convince yourself that GM in more or less all brain areas are negatively
correlated with age, if not corrected for global GM effects, just switch
out the
prop scaling and I'd assume that the contrast then is empty, so I'd agree
with John that the effect is a consequence of the prop scaling (which does
not mean that it contains useful information)
I want to add that I am not sure if in a model that *INCLUDES* total GM there
is an orthogonality issue (age / total GM), if you declare one of the
two as the "regressor of interest". Maybe also check the list on this
question.
Best wishes,
Philipp
Max Planck Institute of Psychiatry
NMR Research Group
Kraepelinstr. 2-10
80804 Munich
Mail: [log in to unmask]
Phone: 0049-89-30622-413
|