search

nancynewlin-masi / MidRISH

Code for implementing MidRISH!
0 stars 0 forks source link

harmonizing more than two datasets #1

Closed johnaeanderson closed 1 week ago

johnaeanderson commented 2 weeks ago

Thank you for this very thorough set of code and for the approach - I agree that selecting a midpoint makes more sense than choosing one site as a reference. My question is what happens if we want to harmonize multiple sites? Should this "just work" assuming we have matched the individuals and that there is equal representation from each site for the template creation?

nancynewlin-masi commented 1 week ago

Hello, thank you for your interest in this project! Moving from the N=2 case to N>2 is a bit more complicated to code. We included a formula for these cases in our paper (formula attached), however we did not make the associated code implementation for N>2. But yes, that is the idea; We would adjust the formula from k=2 to k=N and include representation from all sites and compute some "center" or "mid" space that is the average of each site's coefficients.

image

Link: https://www.sciencedirect.com/science/article/abs/pii/S0730725X24000894

johnaeanderson commented 1 week ago

Thanks!