search

andersonwinkler / PALM

PALM: Permutation Analysis of Linear Models
69 stars 28 forks source link

comparing correlations of MRI metrics with clinical/behavioural scores between groups #23

Open themostgod opened 4 years ago

themostgod commented 4 years ago

I haven't really found an option to compare two independent (healthy controls vs. patients) correlation coefficients of MRI metrics (e.g. FA) with neuropsychological measures - i.e. does FA correlate with IQ more in this group than in the other? I know one can still go for Fisher’s r-to-z approach after computing the r values, but is PALM able to do it more "elegantly"? Thank you!

P.S. I love the software, really thanks for all the work!

andersonwinkler commented 4 years ago

Hi @themostgod, it can be done in PALM if you reformulate the problem as an interaction (which is what it is, I think).

The design matrix would have 4 regressors:

EV1: IQ for group 1, zero for group 2 EV2: IQ for group 2, zero for group 1 EV3: intercept (ones) for group 1, zero for group 2 EV4: intercept (ones) for group 2, zero for group 1

The contrasts are then: C1: [1 -1 0 0] C2: [-1 1 0 0]

You'd get a test statistic and a p-value; the latter is the same as for the difference in correlations, corrected for multiple testing via permutations.

Hope this helps!

All the best,

Anderson