LOGISTIC REGRESSION FOR EMPIRICAL STUDIES OF MULTIVARIATE SELECTION.

Area of Science:

• Evolutionary biology
• Quantitative genetics
• Statistical modeling

Background:

• Adaptive evolution studies require understanding genetic bases and selection pressures.
• Multivariate selection analyses commonly use multiple linear regression.
• Linear regression may have limitations with nonnormal traits and dichotomous fitness data.

Purpose of the Study:

• To propose logistic regression as a superior statistical method for estimating multivariate selection.
• To demonstrate the utility of logistic regression in adaptive evolution research.
• To provide a statistically robust alternative to linear regression for selection studies.

Main Methods:

• Applied logistic regression to analyze multivariate selection.
• Compared logistic regression with traditional linear regression approaches.
• Transformed logistic regression estimates for direct use in microevolutionary models.

Main Results:

• Logistic regression is more appropriate for dichotomous fitness outcomes than linear regression.
• Estimates from logistic regression can be readily integrated into adaptive evolution equations.
• The methodology proved effective on two existing biological datasets.

Conclusions:

• Logistic regression provides a more suitable framework for analyzing multivariate selection.
• This method enhances the accuracy of selection strength estimation in evolutionary studies.
• Widespread adoption of logistic regression is recommended for empirical selection research.