Abstract

YANEZ CANAL, S; LOPERA GOMEZ, C. M  and  JARAMILLO ELORZA, M. C. Statistical Methods for Competing Risks: A Comparative Study. Ciencia en Desarrollo [online]. 2014, vol.5, n.2, pp.87-97. ISSN 0121-7488.

The failure time of a system with two failure modes can be modeled as a series system or a competing risks model. Each unit has a potential failure time associated with each failure mode; the observed failure time is the minimum of such potential times. If the competing risks are ignored and there is only one event of interest, the cause-specific hazard is used, which assumes that the other risks does not exist. This is the setting for the classical methods of Kaplan-Meier, log-rank test and Cox proportional hazards model, therefore their results answer questions related to the "pure" effect associated to a single cause. On the other hand if the competing risks are considered as working jointly, the subhazard function or the hazard of the subdistribution are used. Thus, alternative methodologies have been developed to take into account the latter situation. Such as, the cumulative incidence function (CIF), Gray's test and Pepe and Mori's test, and finally competing risks regression, as the counterpart of Cox model. In this paper these methods are showed and interpreted. They are illustrated with a real case in Interconexión Eléctrica S.A. E.S.P. (ISA).

Keywords : Competing risks regression; Cox's regression model; Cumulative incidence function (CIF); Gray's test; Kaplan-Meier; Log-rank test; Pepe & Mori's test.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )