fbst: An R package for the Full Bayesian Significance Test for testing a sharp null hypothesis against its...

Area of Science:

• Psychology and Cognitive Sciences
• Statistical Modeling
• Bayesian Inference

Background:

• Null hypothesis significance testing (NHST) and p-values are standard in psychology but face criticism.
• Existing alternatives to NHST are limited, creating a need for new methods.
• The Full Bayesian Significance Test (FBST) has theoretical and practical advantages as a Bayesian alternative.

Purpose of the Study:

• Introduce the fbst R package for implementing the Full Bayesian Significance Test (FBST).
• Demonstrate FBST as a Bayesian alternative to NHST and p-values.
• Provide practical examples of FBST application in cognitive science research.

Main Methods:

• Implementation of the Full Bayesian Significance Test (FBST) using the fbst R package.
• Calculation of the e-value, representing Bayesian evidence against a sharp null hypothesis.
• Application to any Bayesian model with obtainable posterior distributions.

Main Results:

• The fbst package provides the e-value for hypothesis testing.
• It allows computation of asymptotic p-values and generates visualizations for interpretation.
• Demonstrated practical application of FBST in three common cognitive science statistical procedures.

Conclusions:

• The fbst R package offers a valuable Bayesian alternative for hypothesis testing.
• FBST provides theoretical and practical benefits over traditional NHST.
• Encourages wider adoption of FBST in cognitive science research for sharp null hypothesis testing.