Title:A Generalized Framework for Approximate Co-Sufficient Sampling

Abstract:Approximate co-sufficient sampling (aCSS) offers a principled route to hypothesis testing when null distributions are unknown, yet current implementations are confined to maximum likelihood estimators with smooth or linear regularization and provide little theoretical insight into power. We present a generalized framework that widens the scope of the aCSS method to embrace nonlinear regularization, such as group lasso and nonconvex penalties, as well as robust and nonparametric estimators. Moreover, we introduce a weighted sampling scheme for enhanced flexibility and propose a generalized aCSS framework that unifies existing conditional sampling methods. Our theoretical analysis rigorously establishes validity and, for the first time, characterizes the power optimality of aCSS procedures in certain high-dimensional settings.