Related Experiment Videos
A semiparametric model for the analysis of recurrent-event panel data
1Simon Fraser University, Burnaby, British Columbia, Canada. rbalshaw@stat.sfu.ca
Biometrics
|June 20, 2002
Summary
This study introduces a robust semiparametric model for recurrent-event data, particularly when events are costly to detect. The model offers efficient estimation and diagnostics for analyzing phenomena like cancer recurrences.
Area of Science:
- Biostatistics
- Epidemiology
- Statistical Modeling
Background:
- Longitudinal studies often track nonterminating or recurring phenomena.
- Recurrent-event data arises from events like epileptic seizures or cancers.
- Detecting recurring events can be expensive or invasive, limiting data availability.
Purpose of the Study:
- To present a semiparametric model for recurrent-event data with costly detection.
- To account for overdispersion in count data using subject-specific effects.
- To provide robust and efficient estimation methods.
Main Methods:
- A multiplicative intensity model with a flexible nonparametric baseline intensity function.
- Inclusion of a random subject-specific effect to handle overdispersion.
- Utilizing quasi-likelihood estimating functions for robust estimation.
Main Results:
- The semiparametric estimators are highly efficient compared to parametric methods.
- The method provides diagnostics for testing parametric models, including baseline intensity functions.
- Simulation studies confirm appropriate small-sample characteristics.
Conclusions:
- The proposed semiparametric model offers a robust and efficient approach for analyzing recurrent-event data.
- The method is applicable to various fields, including medical research and ecological studies.
- The model's flexibility and diagnostic capabilities enhance its practical utility.