Improving inferential analyses predata and postdata

Area of Science:

• Psychology
• Statistics

Background:

• Traditional research methodology involves pre-data power analyses and post-data significance testing.
• Null hypothesis significance tests offer limited information and are prone to misuse.
• Alternative post-data methods can provide more useful information regarding probabilistic outcomes.

Purpose of the Study:

• To propose a revised two-step statistical procedure for researchers.
• To advocate for a shift from traditional significance testing to probabilistic outcome estimation.
• To introduce the a priori procedure as a replacement for conventional power analysis.

Main Methods:

• The study suggests replacing conventional power analysis with an a priori procedure focused on parameter estimation.
• It proposes estimating probabilities of being better or worse off, depending on treatment, as a post-data analysis.
• This approach is based on the work of Trafimow and colleagues.

Main Results:

• The proposed method offers a higher grade of useful information compared to significance testing.
• It allows for the estimation of probabilistic advantages or disadvantages associated with different outcomes.
• This facilitates a more nuanced understanding of treatment effects.

Conclusions:

• The conventional two-step statistical process (power analysis and significance testing) should be replaced.
• A new two-step procedure is recommended: a priori parameter estimation followed by post-data probabilistic outcome assessment.
• This revised approach enhances the utility and interpretability of research findings.