Abstract
In this paper, the structure of semicompeting risks data, dened by Fine, Jiang & Chappell (2001), is studied. Two events are of interest: a nonterminal and a terminal event, the last one, can censor the non-terminal event, but not vice versa. Due to the possible dependence between the times until the occurrence of such events, two approaches are evaluated: modelling the bivariate survival function through Archimedean copulas and a shared frailty model. A simulation is conducted to examine its performance and both approaches are applied to a real data set of patients with chronic kidney disease (CKD).
Highlights
Chronic kidney disease (CKD) is considered a worldwide problem and a great increase of deaths caused by chronic kidney disease (CKD) has been observed from 1990 to 2010 (Martín-Cleary & Ortiz 2014)
We are concerned with the analysis of a dataset with end-stage renal disease (ESRD), stages IV and V, of CKD patients, followed from 2009 to 2011, from ve cities in Colombia: Manizales, Monteria, Rionegro, Sincelejo and Tunja
Increased attention has been given to semicompeting risks data, and there are two main diculties in addition to the existence of censored data: (i) the hypothesis that we have a pair of random variables with the restriction that one of them is always smaller than the other one or one of them is not observed and (ii) the time until the non-terminal event and time until the terminal event cannot be considered as independent random variables
Summary
Chronic kidney disease (CKD) is considered a worldwide problem and a great increase of deaths caused by CKD has been observed from 1990 to 2010 (Martín-Cleary & Ortiz 2014). Increased attention has been given to semicompeting risks data, and there are two main diculties in addition to the existence of censored data: (i) the hypothesis that we have a pair of random variables with the restriction that one of them is always smaller than the other one (progression necessarily occurs before death) or one of them is not observed and (ii) the time until the non-terminal event and time until the terminal event cannot be considered as independent random variables This possible dependence must be taken into account into statistical analysis. We focus on a three-state process, known as Illness-Death process, initially studied by Fix & Neyman (1951) and proposed by Xu, Kalbeisch & Tai (2010) for semicompeting risks data
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
