| Summary: | 博士 === 國立成功大學 === 統計學系碩博士班 === 98 === Current status data arise naturally from tumorigenicity experiment, biomedicine, econometrics, demographic and sociology studies. Moreover, clustered current status data often occur when animals are from same litter in tumorigenicity experiments. The only information extracted from current status data is that the true survival times are before or after the monitoring times. Consequently, the nonparametric maximum likelihood estimator of survival function converges at rate to a complicated limiting distribution (Groeneboon and Wellner, 1992). Hence, semiparametric regression and linear regression have been extended for independent current status data to estimate the survival functions whose rate converge at .
However, a straightforward application of these statistical methods to clustered current status data is not appropriate. Therefore, marginal approaches are applied in this dissertation to construct estimating functions for deriving the estimators of regression parameters in additive hazards models with clustered current status data. When the marginal survival times follow an additive hazards model, it is natural to use the partial score function derived by Lin et al. (1998) and the efficient score function derived by Martinussen and Scheike (2002) as estimating functions. Then, the asymptotic properties of the estimators of regression parameters in additive hazards models obtained by marginal approaches will be investigated. The implementation of the methods is illustrated through one dataset.
|
|---|
