4 ms·
Amusing comments provided you are not Vinod. Three sites and I would hope that connectomes were computed interleaved between cases and controls and analyzed wit
by robwwilliams 2y ago
Amusing comments provided you are not Vinod. Three sites and I would hope that connectomes were computed interleaved between cases and controls and analyzed with blinding. Is that standard? Do you think behavioral difference in the scanner could generate false positive connectome differences?
- SubiculumCode 2y agoI am not sure what you mean by interleaved here. In 2013 I am not sure that site differences were as appreciated as they are today. Today, in studies that use multi-site imaging data, ComBat Harmonization is often the method that is employed to account for site effects in two ways: (1) it models site-specific scaling factors and (2) it uses empirical Bayes to improve the estimation of the site parameters for small sample sizes. https://pmc.ncbi.nlm.nih.gov/articles/PMC5845848/ https://pmc.ncbi.nlm.nih.gov/articles/PMC5845848/
- robwwilliams 2y agoBy interleaved I mean that cases and controls were not each in their own run batches, but ideally run A ,B, A, B, Yes ComBat can some of the batch effect if sample sizes are large enough. But here as you point out they are on thin ice.