Associate Professor, Biostatistics
Logistic Regression and Prediction for Health Data
What You'll Learn
- Understand how binary outcomes arise and know the difference between prevalence, risk ratios, and odds ratios
- Use logistic regression to estimate and interpret the association between one or more predictors and a binary outcome
- Understand the principles for using logistic regression to make predictions and assessing the quality of those predictions
3 Modules
12 Hours
4 hrs per module (approx.)
Rating
About Logistic Regression and Prediction for Health Data
This course introduces learners to the analysis of binary/dichotomous outcomes. Learners will become familiar with fundamental tests for two-group comparisons and statistical inference plus prediction more broadly using logistic regression. They will understand the connection between prevalence, risk ratios, and odds ratios. By the end of this course, learners will be able to understand how binary outcomes arise, how to use R to compare proportions between two groups, how to fit logistic regressions in R, how to make predictions using logistic regression, and how to assess the quality of these predictions. All concepts taught in this course will be covered with multiple modalities: slide-based lectures, guided coding practice with the instructor, and independent but structured exercises.
Skills You'll Gain
- Data Analysis
- Logistic Regression
- Probability Distribution
- Statistical Analysis
- Statistical Programming
What You'll Earn
- Certificate of Completion:
- Certificates of completion acknowledge knowledge acquired upon completion of a non-credit course or program.
Welcome Message
Welcome to Logistic Regression and Prediction for Health Data, the final course in the Data Science for Health Research specialization. This course focuses on analyzing binary outcomes, comparing two groups, and predicting health outcomes using logistic regression in R. Learners will explore prevalence, risk ratios, odds ratios, and evaluate prediction model performance through ROC curves and calibration. Practical exercises, guided coding, and quizzes reinforce learning throughout the course.
This abbreviated syllabus description was created with the help of AI tools and reviewed by staff. The full syllabus is available to those who enroll in the course.
Course Schedule
Module 1: Simple Comparisons of Binary Outcomes
- Video: Data Science for Health Research: Specialization Introduction
- Reading: Meet Your Instructors
- Reading: Welcome & Course Syllabus
- Discussion Prompt: Meet Your Fellow Global Classmates
- Reading: Pre-Course Survey
- Reading: Introduction To and How To Use Independent Guides
- Reading: Introduction to the BPUrban Data
- Video: How and When Binary Outcomes Can Arise
- Video: A Need for Models Beyond Linear Regression
- Video: Binary Outcomes, Comparing Between Two Groups (Part 1)
- Video: Binary Outcomes, Comparing Between Two groups (part 2)
- Video: Binary Outcomes, Comparing Between Two groups (part 3)
- Video: Guided Practice: Z-Test
- Video: Guided Practice: Fisher's Exact Test
- Reading: 1.2 Independent Guide
- Video: Analyzing a Binary Outcome and Binary Exposure with the Odds Ratio
- Video: Interpreting the Odds Ratio
- Video: 2x2 Example: The WCGS Cardiovascular Study
- Graded Assignment: 1.2 Practice Quiz
- Discussion Prompt: 1.2 Discussion Prompt
- Reading: 1.2 Discussion Prompt Suggested Answer
- Discussion Prompt: End of Module 1 Discussion Prompt
- Reading: End of Module 1 Discussion Prompt Suggested Answer
Module 2: Introducing Logistic Regression
- Video: Limitations of the 2x2 Table Analysis
- Video: Logistic Regression: A First Look
- Video: Visualizing and Interpreting a Logistic Regression
- Video: Revising the 2x2 Example: WCGS Cardiovascular Study
- Video: Guided practice: Fitting a Simple Logistic Regression Against One Variable
- Reading: 2.1 Independent Guide
- Graded Assignment: 2.1 Practice Quiz
- Video: Extending the WCGS Cardiovascular Model with Multivariable Logistic Regression
- Video: Prediction with Multivariable Logistic Regression
- Video: Logistic Regression: A Recap and Review
- Video: Guided Practice: Fitting a Logistic Regression Against More Than One Variable
- Video: Guided Practice: Calculating Predicted Probabilities
- Video: Guided Practice: Visualizing a Fitted Logistic Regression Model
- Reading: 2.2 Independent Guide
- Graded Assignment: 2.2 Practice Quiz
Module 3: Assessing Logistic Regression Models
- Video: Why Do We Need to Assess Predictions?
- Video: Extracting Probabilities from a Logistic Regression
- Video: How Do We Determine if Predicted Probabilities are "Good"?
- Video: Model Calibration
- Video: Hosmer-Lemeshow Test
- Video: Model Discrimination
- Video: Changing the Cutpoint Changes Sensitivity and Specificity
- Video: Receiver Operating Characteristic (ROC) Curve
- Video: Area Under the ROC Curve (AUC)
- Video: AUC Example: Risk of Coronary Heart Disease
- Video: Brier Score
- Video: Cross Validation
- Video: Guided Practice: Assessing the Predictive Ability of Logistic Regression Models
- Video: Guided Practice: ROC and AUC
- Video: Guided Practice: Brier Score
- Reading: 3.3 Independent Guide
- Video: Case Study: Treatment of Testicular Cancer
- Graded Assignment: 3.3 Practice Quiz
- Discussion Prompt: End of Module 3 Discussion Prompt
- Reading: End of Module 3 Discussion Prompt Suggested Answer
- Reading: Post-Course Survey
Grading Policy
Course materials are available for self-paced learning. Learners must earn an overall grade of 80% to pass and receive the certificate. The course grade is based on three quizzes worth 33.33% each.
Professor of Epidemiology
Course content developed by U-M faculty and managed by the university. Faculty titles and affiliations are updated periodically.
Intermediate Level
There are no formal requirements to take this course. It is expected that learners have a basic understanding of algebra and probability.
