|
Abstract:
|
In medical research, survival data are typically modeled non- or semiparametrically. Parametric survival modeling, however, can reveal additional insights to clinicians. A parsimonious piecewise exponential model, which is proved to be the UMPU test under certain conditions, is presented. At a given a level, this model can identify whether a shift in the hazard rate occurs and estimates the location(s) of the shifts. The final shift may indicate a near "cure," without having to explicitly estimate such a parameter. Such modeling provides an additional descriptive tool in understanding differences in patient response.
|
