Biomedical Engineering Reference
In-Depth Information
for model misspecification, missing covariates, or misclassified covariates with
interval-censored data. A recent reference on the misclassified covariate is Zeng
et al. (2010), who proposed a likelihood-based approach and showed that it
reduces the bias resulting from the naive use of misclassified covariates.
For the discussion above or in general, the failure event that T represents
is assumed to occur for sure. Sometimes this is not the case as some sub-
jects may never experience the event and they are usually said to be cured.
Another terminology often used in the literature for this is to refer the sub-
jects who will eventually experience the event as susceptible and the others
as nonsusceptible. Also, a model for this situation is usually referred to as a
cure model and the observed data as survival data with long-term survivors.
The objectives of interest for the situation include estimation of the effects
of covariates on cure rates and the survival of susceptible subjects. Several
authors have recently considered the inference on cure models with interval-
censored data. For example, both Kim and Jhun (2008) and Liu and Shen
(2009) developed some maximum likelihood approaches and EM algorithms.
But as commented by these authors, more research is still needed for the cure
model with interval-censored data.
Some discussion was given in Section 1.2 about independent or noninfor-
mative interval-censoring. A natural question is how to carry out regression
analysis of interval-censored data when the censoring is dependent or infor-
mative. For such situations, it is easy to see that one has to make certain
assumptions on the relationship between the failure time T of interest and
censoring variables, and these assumptions are not verifiable in general with-
out extra information. As a consequence, one may have to rely on sensitivity
analysis for inference. Even if one believes that an assumed model is reason-
able or reliable, it may not be easy to develop some inference procedures.
Among others, Zhang et al. (2005) proposed some joint frailty models for the
problem and developed an estimating equation-based approach. However, the
 
Search WWH ::




Custom Search