How statistical challenges … produce unreplicable #science
https://osf.io/preprints/psyarxiv/ekmdf
"…when estimating a #causalEffect from data with a #multilevel structure, it is necessary to allow both the intercepts & effects to vary along each grouping structure. In practice, this is a challenge, 1st because we know of no #textbook that offers this recommendation; 2nd because including additional variance components to a model can make it less stable to fit… 3rd because modeling choices will still arise…"
OSF

I often see researchers confuse the effects they can measure in experiments vs observational studies. This fig from https://doi.org/10.1177/00031224211004187 is my go-to when explaining this:

#stats #causaleffect #statstodon #experiments #DAGs #research