Correcting False Positive Rates in Interaction Model Testing

This corrigendum addresses errors and limitations discovered after the publication of Esarey and Sumner's paper on interaction models.

Correction Overview:

The authors recognized significant issues with their previous work, particularly concerning the overconfidence of the Brambor et al. procedure for separately testing multiple hypotheses.

• The original paper correctly identified this limitation.
• This corrigendum details these errors and introduces a new robust procedure to address them.

Understanding the Problem:

The authors uncovered these issues while developing methods for another unrelated paper, highlighting potential pitfalls in applying established procedures.

• The Brambor et al. procedure may be overly optimistic when testing hypotheses individually.
• Its application is appropriate but limited when testing multiple hypotheses simultaneously.

The Solution:

This update provides a comprehensive description of the limitations related to the Benjamini and Hochberg method for separate hypothesis testing (which was presented in their paper).

• The authors suggest a procedure on pp. 1161–1163 based on this methodology.
• This corrigendum introduces an improved, more robust approach developed during this process.

Revised InteractionTest Software:

The article includes updated methods designed to better handle false positives in interaction model analyses.