[BUG] weighted SMD computation both pooled and federated is flawed

owkin / fedeca

Code repository associated with the FedECA article

https://owkin.github.io/fedeca/

Other

7 stars 2 forks source link

Closed jeandut closed 3 months ago

jeandut commented 3 months ago

We are computing traditional SMD on weighted samples which is different from weighted SMD (see i.e. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4626409/)

jeandut commented 3 months ago

Solved partially by #50