CMStatistics 2016: Start Registration
View Submission - CMStatistics
B1179
Title: Confidence bands for a recurrent quantile as a function of the covariates in recurrent event models Authors:  Akim Adekpedjou - Missouri University of Science and Technology (United States) [presenting]
Abstract: The main driver is the need to construct confidence bands for quantile as a function of covariates with recurrent event models. Consider the quantile of the time between occurrences of a unit subject to a recurring event and let $X$ be the possibly time varying covariates associated with the unit. Of interest is the construction of various confidence bands wherein the covariates $X$ and interventions performed after a recurrence are incorporated via a general Cox-type model for recurrent events. We propose three type of bands: (i) the first based on the original asymptotic properties of the properly standardized quantile which are intractable; (ii) the second based on the Khmaladze transformation of the distribution in (i) leading parameter free distribution and lastly (iii) the bootstrap based bands. We present various asymptotic properties of the properly standardized quantile. Small sample properties of the results are examined via a simulation study and illustration of the methods with a real recurrent event data is presented.