Logistic Regression Models

Abstract

This thesis explores the application of logistic regression models-binary, multinomial, and ordinal-for analyzing categorical outcomes in real-world scenarios. It begins with foundational concepts of linear regression and transitions into logistic regression, emphasizing maximum likelihood estimation (MLE) for parameter estimation. Key statistical tools like odds ratios, log-odds, and confidence intervals are interpreted to reveal predictor-outcome relationships. Practical applications include predicting diabetes diagnosis, classifying student program choices, and assessing maternal mortality risks, evaluated using confusion matrix metrics and diagnostic tests. The study highlights logistic regression's efficiency as both a predictive and explanatory tool, offering actionable insights for decision-making in healthcare and education