Learning Objectives

• identify situations in which the generalized linear model (GLM) should be used for analysis
• fit and interpret results of GLM’s for binary and count data
• understand shrinkage and when hierarchical models may be useful in estimation
• fit, interpret, and illustrate results from hierarchical models

Case Study Goals

The goal of this case study is to examine county-specific infant mortality rates as a function of race/ethnicity.

Data

The data from this case study are simulated data (based on real data but generated randomly to eliminate risk of disclosure). The background is that a large HMO wants to know what patient and physician factors are most related to whether a patient’s lung cancer goes into remission after treatment as part of a larger study of treatment outcomes and quality of life in patients with lung cancer.

• Simulated lung cancer data

Resources

• Logistic Regression in R
• Introduction to Multilevel Models in R
• Gelman and Hill, Data Analysis Using Regression and Multilevel/Hierarchical Models
• Introduction to GLMMs
• Details on Example from IDRE

Reports

• Create an interactive visualization in Tableau to show how infant mortality varies across counties, race/ethnicity groups, and time in NC. Visualizations will be presented in class on October 26. This visualization will be included in your final report.

• In a final report due November 14, examine the factors related to infant mortality rates and obtain estimated infant mortality rates for each county, year, and race/ethnicity group using a multilevel model. Write out the model you used in mathematical notation, and incorporate your estimated rates in a Tableau visualization for comparison with that you generated in the interim report. This report is not to exceed 5 pages. Reproducible code for fitting the model should be uploaded along with the report.

The case study grade will be based on the final report.

Slides

• Linear multilevel models

• Multilevel logistic regression