Adaptive Monte Carlo Multiple Testing via Multi-Armed Bandits
Authors: Martin Zhang, James Zou, David Tse
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | On a Parkinson GWAS dataset, the algorithm reduces the running time from 2 months for full MC to an hour. |
| Researcher Affiliation | Collaboration | 1Department of Electrical Engineering, Stanford University 2Department of Biomedical Data Science, Stanford University 3Chan-Zuckerberg Biohub. |
| Pseudocode | Yes | Algorithm 1 The AMT algorithm. |
| Open Source Code | Yes | Code availability. The software is available at https://github.com/martinjzhang/AMT |
| Open Datasets | Yes | We consider a GWAS dataset that aims to identify genetic variants associated with Parkinson s disease (Fung et al., 2006) |
| Dataset Splits | No | The paper does not explicitly mention training, validation, and testing splits for its datasets, nor does it provide details on how the data was partitioned for model development and evaluation in distinct phases. |
| Hardware Specification | Yes | This experiment is run on 32 cores (AMD Opteron TM Processor 6378). |
| Software Dependencies | No | The paper mentions specific statistical methods (e.g., Pearson's Chi-squared test, Agresti-Coull confidence interval) and a GWAS package (PLINK), but it does not specify any software names with version numbers required for reproduction. |
| Experiment Setup | Yes | In the actual implementation, we chose h1 = 100 and γ = 1.1 for all experiments. In the default setting, we consider m=1000 hypothesis tests, out of which 200 are true alternatives... The number of f MC samples per hypothesis is set to be n=10,000 while the nominal FDR is α=0.1. |