This Friday, 3/25 at 12:00 noon Eastern, the International Methods Colloquium will host a talk by Fred Boehmke (University of Iowa) titled “Proper Specification of Non-proportional Hazards Corrections in Duration Models” (a co-authored project with Shuai Jin).
Tha abstract for this presentation follows:
Parametric and nonparametric duration models assume proportional hazards, which means the effect of a covariate on the hazard rate is constant over time. Researchers have developed techniques to test and correct non-proportional hazards, including interacting the covariates with some function of time. Including this interaction term means that the specification now involves time-varying covariates, and the model specification should reflect this feature. However, in situations with no time-varying covariates before the correction, researchers continue to models the duration with only time-invariant covariates. This error results in biased estimates, particularly for the covariates interacted with time. We investigate this issue in over forty Political Science articles and find that of those studies that begin with time invariant covariates and correct for non-proportional hazards the majority suffer from incorrect model specifications. Proper estimation usually produces substantively or statistically different results.
To tune in to the presentation and participate in the discussion after the talk, visit http://www.methods-colloquium.com/ and click “Watch Now!” on the day of the talk. To register for the talk in advance, click here:
The IMC uses GoToWebinar, which is free to use for listeners and works on PCs, Macs, and iOS and Android tablets and phones. You can be a part of the talk from anywhere around the world with access to the Internet. The presentation and Q&A will last for a total of one hour.