Fall Seminar Series - December 4, 2003
University of Minnesota
School of Statistics
College of Liberal Arts
Analyzing Panel Count Data with Informative Observation Time
Chiung-Yu Huang
Division of Biostatistics
University of Minnesota
Thursday,
December 4, 2003
4:00 PM, 115
Ford Hall
Minneapolis,
East Bank Campus
Social
at 3:30 PM, 300
Ford Hall
Abstract
In
many longitudinal studies, observations on recurrent events of study subjects
are taken at several distinct time points, and, instead of recording the
exact event times, only the number of events that have occurred before each
observation time points is known. Data of this type are commonly referred
to as panel count data. In this paper, we assume proportional rate models
for the recurrent event process, where the form of the baseline occurrence
rate function is left unspecified and a subject-specific latent variable acts
multiplicatively in the rate function. The proposed models allow the recurrent
event process and observation times to be correlated through their connections
with the latent variable; furthermore, a specific feature of the model is
that the distributions of the latent variable and the observation times are
not parameterized. Estimation procedures that require no information about
the latent variable are proposed for the cumulative baseline rate function
and the regression parameters. An application to a bladder tumor study is
presented to illustrate proposed methods.
This is joint work with Mei-Chang Wang (Johns Hopkins University) and Ying
Zhang (University of Central Florida)