| Session Title: Using Multilevel Discrete-time Survival Models to Predict Whether and When Events Occur |
| Demonstration Session 586 to be held in Calhoun Room on Friday, November 9, 11:15 AM to 12:00 PM |
| Sponsored by the Quantitative Methods: Theory and Design TIG |
| Presenter(s): |
| Steven Pierce, Michigan State University, pierces1@msu.edu |
| Abstract: The session will introduce the audience to discrete-time survival analysis, which builds on logistic regression to predict not only whether an individual will experience an event, but also when the individual experiences the event. The session will then illustrate how to extend the technique to handle the multilevel case where individuals are drawn from a clustered sample and both individual-level and cluster-level characteristics are used to predict the occurrence of the event. The example data examine whether a door-to-door outreach effort affected whether and when households drawn from 52 different neighborhoods in a small city responded to a survey about neighborhood conditions. The session will cover the data needed to do these models, the concept of censoring, exploratory analysis, software tools, testing multilevel survival models, interpreting output, graphical methods for displaying the results, and recommended resources for those who want to learn more about the technique. |