Hi,
I did a multicenter fixel based analysis, described here: https://elifesciences.org/articles/82088 (code: GitHub - smeisler/Meisler_Reading_FBA: Code used in the Meisler and Gabrieli 2022 eLife paper on fixel based analyses of reading abilities.).
One extra thing you have to do is to average fiber response functions within each center, NOT across everyone. I additionally added a linear regressor for site in my final model to capture remaining variance, but there are probably better ways of accounting for site in the final model.
Best,
Steven