Improving Inferences About Null Effects With Bayes Factors and Equivalence Tests

Area of Science:

• Statistics
• Gerontology

Background:

• Null-hypothesis significance testing (NHST) is often misinterpreted, leading researchers to incorrectly conclude no effect exists when p values are nonsignificant.
• This misinterpretation can hinder accurate scientific conclusions and the advancement of research.

Purpose of the Study:

• To introduce and explain equivalence testing and Bayes factors as statistically sound methods for evaluating the absence of effects.
• To demonstrate the application of these methods using examples from gerontology research.
• To guide researchers in designing studies that can support null or absence-of-effect hypotheses.

Main Methods:

• Equivalence testing: A frequentist statistical approach to determine if an effect is practically negligible.
• Bayes factors: A Bayesian statistical approach to quantify the evidence for or against a null hypothesis compared to an alternative hypothesis.
• Application to gerontology literature: Four case studies illustrating the use of these methods to reject meaningful effects.

Main Results:

• Nonsignificant p values from traditional tests do not logically or statistically support the absence of an effect.
• Equivalence tests and Bayes factors provide robust statistical tools to quantify evidence for the absence of a meaningful effect.
• These methods allow for the falsification of predictions and the quantification of support for null effects.

Conclusions:

• Researchers should adopt equivalence testing and Bayes factors to avoid misinterpreting nonsignificant results.
• These advanced statistical techniques enhance the rigor and accuracy of scientific inferences, particularly in gerontology.
• Implementing these methods leads to more informative studies and reliable conclusions about the presence or absence of effects.