| Summary: | 博士 === 淡江大學 === 數學學系博士班 === 103 === The need for analyzing time-to-event data can arise in a number of applied fields, such as medicine, biology, public health, epidemiology, engineering, economics and demography. A common feature of such data sets is that the event time may not be known completely due to censorings or truncations. In current status data, the event time is not observed directly and is only known to lies before some examining time or not. We consider the estimation problems for current status data under the assumption of additive hazards models when covariates are subject to homogeneous measurement errors.
We proposed to adopt the point of view from Lin, Oakes and Ying(1998) to transform the problem to a Cox proportional hazard model with right censored data. Nevertheless, the measurement errors in “covariates” become heterogeneous after transform. Some modifications were then developed to accommodate such heterogeneous errors for conventional analyzing methods that include corrected score, conditional score and a newly developed method--the extensively corrected score. Our proposal is shown to perform well in simulation study and is applied to diabetes survey data as an illustration of implementation.
|
|---|
