RT @GuidoBiele: @noelTbrewer @dggoldst I think I did not express myself clear enough. My view is that if one wants to claim that results ar…
@noelTbrewer @dggoldst I think I did not express myself clear enough. My view is that if one wants to claim that results are "the same" across samples one needs to run something like an equivalence test or ROPE procedure to support such a claim. https://t.
@KouMurayama @dingding_peng Weights work generally (prov. suff. N) can produce noisier estimates. What method "works" depends on DGP: If bias is due to conditioning on a collider, weighting it is, if it is due to effect modification, standardization/post
@brettjgall @LuHaidong @RexDouglass I fear this is too short for me to understand. Do you mean cond. on a coll. opens a path that leads to selection bias? I agree (https://t.co/4qEonepD5C ). I tried to say is that if one defines biases by from DGPs, then