Reproducibility Package for "Significance-Classification Instability Beyond p-Value Variability: A Monte Carlo Study of Effect Direction and Effect-Estimate Compatibility"

This repository contains the complete computational reproducibility package for the study “Significance-Classification Instability Beyond p-Value Variability: A Monte Carlo Study of Effect Direction and Effect-Estimate Compatibility.” The study uses Monte Carlo simulation to examine the relationship...

Descripción completa

Guardado en:
Autores principales: Evergisto Sare Lara, Juan Francisco Agreda Vega, L. Montoya, Jimmy Aurelio Rosales-Huamaní, Ángel Cárdenas-García, Magaly Tintaya Quispe, Jose Percy Tintaya Quispe
Formato: Software
Lenguaje:inglés
Publicado: 2026
Materias:
direction of effect
statistical significance
Monte Carlo simulation
reproducibility
Meta-analysis and systematic reviews
Statistical Methods and Bayesian Inference
Simulation Techniques and Applications
Acceso en línea:Enlace permanente (DOI o Handle)
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
Sumario:This repository contains the complete computational reproducibility package for the study “Significance-Classification Instability Beyond p-Value Variability: A Monte Carlo Study of Effect Direction and Effect-Estimate Compatibility.” The study uses Monte Carlo simulation to examine the relationship between statistical-significance classification and the stability of effect direction, magnitude, and precision across independent studies generated under the same population mechanism. The reproducibility package includes the R source code, simulation functions, scenario definitions, random-number streams, aggregated simulation results, conditional analyses, tables, figures, metadata, and session information required to reproduce the reported results. The simulation design comprises 509 unique scenarios: 486 factorial scenarios and 25 directed transition-region scenarios, with two overlapping conditions. The factorial scenarios were evaluated using 100,000 paired replications per scenario, while the directed scenarios used 250,000 paired replications. Statistical inference was based on two-sided Welch tests with alpha = 0.05. Random-number generation used the L'Ecuyer-CMRG generator with master seed 20260927. The package also includes a reproducibility verification workflow. Re-execution of the 25 directed scenarios reproduced the original conditional results to machine precision, with a maximum absolute difference of approximately 4.58 × 10^-16. The README file provides the recommended execution order and a description of the repository structure.

Ejemplares similares