Predicting Binary Outcomes: Application to Health

This project introduces students to two common approaches for modeling binary outcomes: the linear probability model and logistic regression. Organizations often seek to predict whether individuals will experience health outcomes that affect healthcare costs and workplace productivity. For example, employers may be interested in predicting whether an employee has high blood pressure. Selecting an appropriate statistical model is important because predictions should be both interpretable and realistic.

Students will work with real-world health screening data from the Illinois Workplace Wellness Study using R. Logistic regression is widely used in practice, but the linear probability model remains popular because of its simplicity and ease of interpretation. Understanding the similarities and differences between these two approaches helps students evaluate the trade-offs between interpretability and predictive accuracy when modeling binary outcomes.

By the end of this project, students will fit both models and compare their predictions for high blood pressure, analyze factors associated with the condition, and gain practical insight into an important econometric and data analytics question.