cit: hypothesis testing software for mediation analysis in genomic applications.

Area of Science:

• Biostatistics
• Computational Biology
• Epidemiology

Background:

• Causal inference methods face challenges with assumption satisfaction, software limitations, and low power in multiple testing.
• Existing methods often rely on estimation rather than hypothesis testing, complicating assumption evaluation.

Purpose of the Study:

• Introduce the causal inference test (CIT) R package for hypothesis testing in causal inference.
• Provide a user-friendly software tool to evaluate statistical significance for potential mediators.

Main Methods:

• The CIT package employs hypothesis testing, allowing evaluation of testable assumptions for statistical significance.
• It supports single and multiple binary/continuous instrumental variables, binary/continuous outcomes, and adjustment covariates.
• A non-parametric permutation-based False Discovery Rate (FDR) option is available for robust estimation.

Main Results:

• Simulation studies validate the CIT package's performance.
• The permutation-based FDR demonstrates a substantial advantage over common multiple testing strategies.
• The software provides P-values and optionally FDR estimates (q-values) for mediators.

Conclusions:

• The CIT package offers a valid and powerful approach to causal inference.
• Permutation-based FDR in CIT outperforms traditional multiple testing methods.
• The open-source R package enhances accessibility for researchers in the field.