Tutorial: Power analyses for interaction effects in cross-sectional regressions

Area of Science:

• Statistics
• Psychometrics
• Quantitative Psychology

Background:

• Interaction analyses, also known as moderation or moderated multiple regression, assess how the relationship between two variables changes based on a third variable.
• Performing accurate power analyses for interactions is challenging, especially with correlated and continuous variables, and existing software often lacks flexibility.
• Key factors influencing statistical power, such as main effects, their correlations, and variable reliability, are not always clearly incorporated into power analyses.

Purpose of the Study:

• To introduce the R package InteractionPoweR and its associated Shiny apps for conducting power analyses of interaction effects.
• To provide researchers with a user-friendly tool for both analytic and simulation-based power analyses, requiring minimal programming experience.
• To demonstrate how factors like main effects, correlations, reliability, and variable distributions impact statistical power for interaction analyses.

Main Methods:

• Utilizing the R package InteractionPoweR for power analyses of interaction effects.
• Employing both analytic and simulation-based approaches within the package.
• Demonstrating the use of parameters such as Pearson's correlation, sample size, reliability, and variable distribution (e.g., binary, Likert scale).

Main Results:

• The InteractionPoweR package facilitates power analyses for interaction effects, accommodating correlated and continuous variables.
• The tutorial illustrates how main effects, their correlations, variable reliability, and distributions influence statistical power.
• The package allows for flexible incorporation of various parameters to enhance the accuracy of power estimations.

Conclusions:

• The R package InteractionPoweR offers a valuable and accessible tool for researchers to conduct robust power analyses for interaction effects.
• Understanding the impact of main effects, correlations, and reliability is crucial for accurate power analysis in moderated regression.
• The package empowers researchers to better plan studies and interpret findings involving interaction effects in statistical models.