Conducting Bayesian-Classical Hybrid Power Analysis with R Package Hybridpower

Area of Science:

• Statistics
• Statistical Modeling

Background:

• Traditional power analysis often overlooks uncertainty, leading to potentially inaccurate results.
• The issue of local optimality in classical power calculations assumes perfect input accuracy, which is rarely true in practice.

Approach:

• This research reviews methods for incorporating uncertainty in power analysis.
• It highlights the Bayesian-classical hybrid approach, implemented in the R package hybridpower.
• The study details how to elicit subjective priors and determine sample size using this hybrid method.

Key Points:

• hybridpower computes both classical and Bayesian-classical hybrid power for common statistical tests (t-test, correlation, regression, ANOVA, sign test).
• The Bayesian-classical hybrid approach provides more realistic power calculations by accounting for local optimality.
• RShiny applications are provided to enhance accessibility for users unfamiliar with R.

Conclusions:

• The Bayesian-classical hybrid approach offers a more robust alternative to classical power analysis by integrating uncertainty.
• The hybridpower package and associated RShiny tools facilitate the application of this advanced method.
• Future developments aim to further expand the capabilities of hybridpower for power analysis.