Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..

Resolution of Simpson's paradox via the common cause principle

Authors: Arshak Hovhannisyan, Armen Allahverdyan

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our study revisits the paradox by re-estimating its frequency with an unbiased data generation process and reaffirms that it is not an artifact of deficient data collection. ... Our numerical result is that the frequency of two inequalities in (52) is 4.29% 0.001%. For this precision it was sufficient to generate N = 107 samples from (48, 50) with n = 8.
Researcher Affiliation Academia Arshak Hovhannisyan Alikhanyan National Laboratory, Yerevan, Armenia EMAIL Armen Allahverdyan Alikhanyan National Laboratory, Yerevan, Armenia EMAIL
Pseudocode No The paper describes theoretical models and mathematical derivations, but does not contain any explicitly labeled pseudocode or algorithm blocks.
Open Source Code No The paper states in its NeurIPS checklist: "Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [NA] Justification: As we explained above, the manuscript does not present new experimental results."
Open Datasets Yes Section 4 considers published data from Ref. [16] on a case of smoking and surviving. ... Section 5 treats data on COVID-19, which was suggested in Ref. [31].
Dataset Splits No The paper re-analyzes existing datasets and performs numerical simulations, but does not specify training/test/validation splits in the traditional sense. For the smoking data, it performs coarse-graining of existing age groups: "Now B = {b, b}, b = B1 B2 B3 B4 B5, b = B6" but this is not a train/test split.
Hardware Specification No The paper describes theoretical and numerical analysis without mentioning specific hardware used for computations or simulations.
Software Dependencies No The paper does not specify any software libraries, tools, or their version numbers used for the numerical results or analysis.
Experiment Setup Yes For modeling a non-informative Dirichlet density we find it natural to take α1 = ...αn = 1/n. ... Our numerical result is that the frequency of two inequalities in (52) is 4.29% 0.001%. For this precision it was sufficient to generate N = 107 samples from (48, 50) with n = 8.