Evaluating the discriminatory power of a multiple logistic regression model

Area of Science:

• Biostatistics
• Epidemiology
• Medical Statistics

Background:

• Assessing the goodness-of-fit for multiple logistic regression (MLR) models is crucial in statistical analysis.
• Existing goodness-of-fit measures lack a clear consensus on their suitability.
• The discriminatory power of MLR models needs reliable estimation.

Purpose of the Study:

• To propose a new, simple measure for the discriminatory power of fitted MLR models.
• To compare the proposed measure with existing goodness-of-fit statistics.
• To evaluate the practical utility and interpretation of the new measure.

Main Methods:

• A novel goodness-of-fit index, J*, is proposed, based on the maximization of Youden's J index.
• The performance of J* is compared against several established goodness-of-fit statistics.
• The proposed measure is illustrated using real-world data from the Lipid Research Clinics Prevalence Study.

Main Results:

• The proposed J* index demonstrates effectiveness in assessing MLR model fit.
• J* provides a valuable measure of discriminatory power.
• The measure is shown to be comparable to existing statistics.

Conclusions:

• The J* index is suggested as a useful alternative for evaluating MLR model goodness-of-fit.
• J* offers a simple and practical interpretation for researchers.
• This measure enhances the assessment of logistic regression model performance.