A goodness-of-fit test for the latency in a mixture cure model with covariates
In classical survival analysis, it is assumed that all individuals will eventually experience the event of interest. However, in many situations a subset of subjects never experiences the event and is therefore considered “cured,” with infinite survival time. This phenomenon is addressed using mixture cure models.
Throughout this talk, a general goodness-of-fit test is proposed for the latency in a mixture cure model. In the presence of right censoring and a cure fraction a formal test is constructed to check the validity of three common models for the latency: a fully parametric model, a semiparametric Cox model and an accelerated failure time model. The asymptotic behaviour of the test statistic will be derived and to calibrate the test in practice a bootstrap method is presented. In addition, an extensive simulation study and a real data application will be presented to show the performance of the new proposal in practice.
Palabras clave: Survival; Mixture cure model; Goodness-of-fit; latency.