Title: Modeling recurrent gap times and conditional estimating equations
Authors: Ioana Schiopu-Kratina - University of Ottawa (Canada) [presenting]
Hai Yan Liu - University of Ottawa (Canada)
Mayer Alvo - University of Ottawa (Canada)
Pierre-Jerome Bergeron - Google (Canada)
Abstract: A semiparametric approach to the analysis of data from right censored recurrent events processes is presented. The dependence of the gap time between consecutive events on a set of covariates is explored. While the entire distribution of each gap time is not modeled, a regression-like dependence is specified for their conditional mean and variance. Under certain conditions on censoring, one can construct normalized estimating functions that are asymptotically unbiased and contain only observed data. Based on these estimating functions one can set up appropriate equations, which are a particular instance of generalized estimating equations. Solutions to these equations provide estimators for the regression and over dispersion parameters. Modern mathematical techniques are used to prove the existence, consistency and asymptotic normality of a sequence of such estimators. Examples that illustrate the theoretical approach are given. Numerical examples with simulated data are presented. A comparison of this methodological and technical approach to other comparable work in the field is presented. Conclusions based on the numerical results are also presented.