A Gentle Introduction to the Comparison Between Null Hypothesis Testing and Bayesian Analysis: Reanalysis of Two...

Area of Science:

• Statistics
• Scientific Methodology
• Biostatistics

Background:

• The use and interpretation of P values in scientific discovery have been debated for nearly a century.
• Recent attention, including a statement from the American Statistical Association, highlights concerns regarding P value misuse.
• Current practices often dichotomize scientific evidence and rely heavily on statistical models for inference.

Purpose of the Study:

• To present an alternative approach to analyzing evidence from randomized controlled trials.
• To contrast the traditional null hypothesis significance testing with a Bayesian approach.
• To provide insights for non-statisticians who use statistical models in research.

Main Methods:

• A coin flipping experiment was used to illustrate statistical concepts.
• Real-world data from a randomized controlled trial was reanalyzed.
• The traditional null hypothesis significance testing method was compared with a Bayesian approach.

Main Results:

• The study demonstrates how a Bayesian approach can report evidence without dichotomizing findings.
• Contrasting the two methods highlights the limitations of null hypothesis significance testing in certain contexts.
• The reanalysis provides practical examples of alternative evidence reporting.

Conclusions:

• Reporting evidence directly, as facilitated by a Bayesian approach, offers a valuable alternative to P value dichotomization.
• This method can lead to more nuanced and accurate scientific inference.
• The paper encourages researchers to consider Bayesian methods for a more transparent handling of statistical evidence.