{"id":"arxiv/2611.03255","url":"https://arxiv.org/abs/2611.03255","tags":["stat.ME","math.ST"],"score":0,"title":"Conditional independence testing through marginal permutation","domain":"stat.ME","source":"arxiv","authors":["Jiahao Ai","Ziang Niu"],"reviews":0,"subject":"stat.ME","summary":"Marginal permutation is exact for testing marginal independence but generally invalid for conditional independence because it breaks the dependence between the exposure ($\\bm X$) and covariates ($\\bm Z$). In this paper, we show that empirical studentization can nevertheless make this deliberately misspecified resampling scheme asymptotically valid for important conditional independence tests based on scores. The resulting procedures are novel in the sense that they are valid in finite samples when the exposure is marginally independent of the outcome ($\\bm Y$) and covariates, and…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-11-02","reactions":{}}