A Regularization-Based Adaptive Test for High-Dimensional GLMs

Authors: Chong Wu, Gongjun Xu, Xiaotong Shen, Wei Pan

JMLR 2020 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Via simulations, its superior finite-sample performance is demonstrated over several representative existing methods. In addition, we apply it and other representative tests to an Alzheimer s Disease Neuroimaging Initiative (ADNI) data set, detecting possible gene-gender interactions for Alzheimer s disease. Sections 3 and 4 are titled "Simulations" and "Real data analyses" respectively.
Researcher Affiliation Academia Chong Wu EMAIL Department of Statistics, Florida State University, FL, USA; Gongjun Xu EMAIL Department of Statistics, University of Michigan, MI, USA; Xiaotong Shen EMAIL School of Statistics, University of Minnesota, MN, USA; Wei Pan EMAIL Division of Biostatistics, University of Minnesota, MN, USA.
Pseudocode No The paper describes methods mathematically and in paragraph text but does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes We have released open source R package aispu implementing the proposed test on Git Hub (https://github.com/Chong Wu-Biostat/aispu), and will upload it to CRAN soon.
Open Datasets Yes In addition, we apply it and other representative tests to an Alzheimer s Disease Neuroimaging Initiative (ADNI) data set, detecting possible gene-gender interactions for Alzheimer s disease. See www.adni-info.org for the latest information.
Dataset Splits No The paper mentions generating 1,000 datasets for simulation and sampling n/2 cases and n/2 controls for logistic regression simulation. For real data, it describes subject grouping for outcome definition but does not specify standard training/test/validation splits needed to reproduce experiments.
Hardware Specification No The acknowledgments mention the "Minnesota Supercomputing Institute" as a resource, but no specific hardware components like GPU or CPU models, or memory details are provided.
Software Dependencies No The paper mentions using "pmvnorm() in R package mvrnorm" and that they have released "R package aispu". However, it does not provide specific version numbers for these R packages or for R itself, which is required for reproducibility.
Experiment Setup Yes For each simulation setting, we generated 1,000 data sets to evaluate the empirical size and power at the significance level α = 0.05. The candidate set of γ for the ai SPU was taken to be Γ = {1, 2, . . . , 6, } unless otherwise stated. We set ρ = 0 to generate independent SNPs unless otherwise stated. We set ϑ1 = ϑ2 = 0.4 and other ϑj = 0. We set n = 200, q = 1000, and p = 1000. To account for multiple testing, we applied the Bonferroni correction and used a slightly conservative cutoff0.05/100 = 5 10 4.