{"id":"arxiv/2601.00541","url":"https://arxiv.org/abs/2601.00541","tags":["math.ST"],"score":0,"title":"Asymptotic Distribution-Free Tests for Ultra-high Dimensional Parametric Regressions via Projected Empirical Processes and $p$-value Combination","domain":"math.ST","source":"arxiv","authors":["Falong Tan","Shan Tang","Lixing Zhu"],"reviews":0,"subject":"math.ST","summary":"This paper develops a novel methodology for testing the goodness-of-fit of sparse parametric regression models based on projected empirical processes and p-value combination, where the covariate dimension may substantially exceed the sample size. In such ultra-high dimensional settings, traditional empirical process-based tests often fail due to the curse of dimensionality or their reliance on the asymptotic linearity and normality of parameter estimators--properties that may not hold under ultra-high dimensional scenarios. To overcome these challenges, we first extend the classic martingale…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-01-02","reactions":{}}