Data as evidence

Area of Science:

• Physiology
• Statistical Analysis
• Biostatistics

Background:

• Traditional statistical methods in experimental physiology often rely heavily on P-values.
• Concerns exist regarding the interpretation and potential misuse of P-values for declaring statistical significance.
• Alternative statistical methodologies, such as Bayesian and likelihood approaches, are gaining attention.

Purpose of the Study:

• To introduce and explain the likelihood approach to statistical analysis in experimental physiology.
• To demonstrate the application of the likelihood approach with both artificial and real-world datasets.
• To compare the efficacy of the likelihood approach against traditional methods like P-values and chi-squared tests.

Main Methods:

• Review of P-value usage and limitations in physiological research.
• Introduction to the likelihood approach with a simplified artificial example.
• Application of the likelihood approach to a realistic physiological dataset.
• Comparison of likelihood analysis with the chi-squared test for categorical data.

Main Results:

• The likelihood approach provides a mathematically and statistically accurate framework for data analysis.
• Demonstrated superiority of the likelihood approach over the chi-squared test for categorical data analysis.
• Illustrative examples show the practical application and conceptual clarity of the likelihood method.

Conclusions:

• The likelihood approach offers a robust and accurate statistical methodology for experimental physiology.
• Despite its advantages, the likelihood approach is underrepresented in current scientific literature and training.
• Further adoption of the likelihood approach could enhance the rigor of statistical analysis in physiological studies.