Jfortin1 / ComBatHarmonization

Harmonization of multi-site imaging data with ComBat
268 stars 107 forks source link

Harmonization for multi-study project where each study has multiple sites #42

Open pauldhami opened 1 year ago

pauldhami commented 1 year ago

Greetings,

I am hoping to combine the MRI data from 2 independent studies, each of which had 4 unique sites. Thus, there are in total 8 sites where data was collected that I am interested in aggregating.

I had two questions:

1) In this scenario, because the actual study site (8 in total) is the smallest unit, would the batch id vector be made up of the site information, and not study?

2) In any post-harmonization statistical analyses, for example linear regression, is it suggested/would there be any value in including study (2 in total) as a covariate? On that note, I assume there would be no need for site (which would be already addressed in the initial harmonization procedure) to be included as a covariate as well?

Sorry for what are naive questions I'm sure. Any response would be greatly appreciated! Thank you.