Bayesian alternatives for common null-hypothesis significance tests in psychiatry: a non-technical guide using JASP

Area of Science:

• Statistics
• Psychiatry Research

Background:

• Classical null hypothesis significance testing (NHST) is a popular inferential framework but has inherent restrictions.
• Bayesian analysis offers a complementary approach to NHST but has been underutilized due to limited accessible software.
• JASP is an open-source statistical package providing a graphical interface for both Bayesian and NHST analyses.

Purpose of the Study:

• To provide an applied introduction to Bayesian inference with Bayes factors using JASP.
• To compare and contrast Bayesian alternatives with common classical null hypothesis significance tests.
• To illustrate the strengths and limitations of both NHST and Bayesian hypothesis testing.

Main Methods:

• Utilized JASP software to perform Bayesian analyses alongside traditional NHST.
• Applied Bayesian methods to common statistical tests including correlations, frequency distributions, t-tests, ANCOVAs, and ANOVAs.
• Compared results from Bayesian inference (Bayes factors) and NHST (p-values).

Main Results:

• Bayes factors can effectively complement p-values in hypothesis testing by offering additional inferential information.
• Bayes factors allow for the quantification of relative evidence supporting both the alternative and null hypotheses.
• The magnitude of evidence provided by Bayes factors can be presented as an easily interpretable odds ratio.

Conclusions:

• Bayesian analysis, while not a new method, has been largely inaccessible to many psychiatry researchers.
• JASP software simplifies reproducible Bayesian hypothesis testing through a user-friendly graphical interface.
• The "point and click" environment of JASP is familiar to users of statistical packages like SPSS, facilitating adoption.