Analysis Of Recurrent Event Data With Nonmonotone Missing At Random Covariate And Informative Censoring

Recurrent event is a series of event that occurs in epidemiologic studies,biomedical research and so on.In these fields,individuals will experience the same kind of event for several times,which is called recurrent event.An important feature is that the recurrent times are ordered and correlated.There are two kinds of data,timedependent covariates and time-independent covariates,which should be distinguished while modeling.For the design of experiments and the heterogeneity of individuals,recurrent event data often has censoring.In this article,we consider that the censoring times are informative,which means they can give some information about the recurrent processes.Thus,a frailty model is needed.In addition,the data may have missing,which causes the estimation to be biased.So we need to know the missing patterns and the missing mechanism,and the methods under these patterns and mechanism.This article supposes the recurrent event data to be nonmonotone and missing at random(MAR),and will ultimate the inverse probability weighting(IPW)method to estimate parameters with observed data.This article will construct a multivariate recurrent event surviving model with time-dependent covariates missing by the idea of Zhao et al.(2012).Learning from Sun and Tchegen(2018),we will use IPW method,to estimate time-dependent and timeindependent parameters,and the cumulative hazard function.After that,the proof of consistence and asymptotic normality will be given.