Robust group and simultaneous inferences for high-dimensional single index model

Authors: Weichao Yang, Hongwei Shi, Xu Guo, Changliang Zou

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Numerical results indicate that the new procedures can be highly competitive among existing methods, especially for heavy-tailed errors. (Abstract) In this section, extensive simulation studies are carried out to evaluate the numerical performance of the proposed methods for the global inference problem described in (1.2). (Section 4)
Researcher Affiliation Academia Weichao Yang School of Statistics Beijing Normal University, Beijing, China yangweichao@mail.bnu.edu.cn Hongwei Shi School of Statistics Beijing Normal University, Beijing, China shihongwei21@mail.bnu.edu.cn Xu Guo School of Statistics Beijing Normal University, Beijing, China xustat12@bnu.edu.cn Changliang Zou NITFID, School of Statistics and Data Science, LPMC and KLMDASR and LEBPS, Nankai University, Tianjin, China zoucl@nankai.edu.cn
Pseudocode Yes Our inference procedure is summrized in the Algothrim 1 as follows. (Section 2) Algorithm 2: multiple testing procedure (Appendix A.3)
Open Source Code Yes Code is avaliable in supplementary materials. (NeurIPS Paper Checklist Q4 Justification)
Open Datasets Yes We apply our methods on a dataset about riboflavin (vitamin B2) production rate with Bacillus Subtilis. This dataset is made publicly by [7] and has been analyzed by many authors, for instance [37], [47], [29], and [22]. The dataset riboflavin can be obtained from the R package hdi. (Appendix A.7)
Dataset Splits No The paper mentions training and testing, but does not explicitly state validation splits or cross-validation settings.
Hardware Specification Yes We use Intel(R) Xeon(R) Silver 4208 CPU @ 2.10GHz. (Section 4)
Software Dependencies No We use the R-package ncvreg [6]. (Section 4) Here, we use R package SILM [59] to implement the three-step testing procedure... (Section 4) We use the R function p.adjust(...,method="BH") to implement the BH procedure. (Appendix A.6) None of these specify version numbers for the software.
Experiment Setup Yes We consider n = 200, 500 and p = 800. ... We consider two different error distributions for ϵ which is independent of X: (1) standard normal distribution N(0, 1); (2) Cauchy distribution or equivalently Student s t distribution with 1 degree of freedom, t(1). ... pout of the responses are picked at random and increased by mout-times maximum of original responses, shorted as pout +mout max(responses). (Section 4) We consider the following five different choices for G to test the hypothesis (1.2): G1 = {10, 11, p 4, p 3, p 2, p 1, p}, G2 = {3, . . . , 6} G1, G3 = {10, . . . , 59, p 149, . . . , p}, G4 = {3, . . . , 6} G3. (Section 4)