Uniform TitleSequential analysis of clustered survival data by marginal methods
NameHou, Bo (author), Sackrowitz, Harold (chair), Zhang, Cunhui (internal member), Xie, Minge (internal member), Liu, Mengling (outside member), Ying, Zhiliang (outside member), Rutgers University, Graduate School - New Brunswick,
SubjectStatistics and Biostatistics,
Survival analysis (Biometry)
DescriptionClustered survival data are a type of multivariate survival data with naturally formed clusters so that event times within a cluster are parallel to each other and correlated. Lee, Wei and Amato (1992) introduced a semiparametric model for the analysis of clustered survival data that assumes event times follow a proportional hazards model.
Sequential analysis of clustered survival data arises in clinical studies in which patients are followed over time and interim analyses are performed. This thesis studies the sequential analysis of clustered survival data with staggered patient entry by adapting Lee, Wei and Amato's approach. It is shown that the two-parameter score process converges to a Gaussian random field irrespective of the correlation within clusters and staggered patient entry. The regression parameter estimator obtained at each time point has the desired properties including consistency and asymptotic normality. A consistent estimator of the baseline cumulative hazard function is also given. More importantly, we propose a novel optimal weighting strategy. We show that the resulting score process not only produces a more efficient estimator, but also has the important property of (asymptotically) independent increments. The latter can be used in conjunction with the well-known error-spending functions to construct proper boundaries in group sequential testing. Finally a sample size calculation formula is given for designing clinical
trials with clustered survival time as the endpoint.
NoteIncludes bibliographical references (p. 50-52).
CollectionGraduate School - New Brunswick Electronic Theses and Dissertations
Organization NameRutgers, The State University of New Jersey
RightsThe author owns the copyright to this work.