Title: Nonparametric hypothesis testing for isotonic survival models with clustering
Authors: John Eustaquio - University of the Philppines - Diliman (Philippines) [presenting]
Abstract: Nonparametric hypothesis testing procedures based on the bootstrap were developed in testing for constant clustering effect in a survival model that incorporates the clustering effect into the Cox Proportional Hazards model. In a clustered survival model, bootstrap estimators of the cluster-specific parameters are consistent. Simulation studies indicate that the procedure is correctly-sized and powerful in a reasonably wide range of data. The test procedure for constant cluster effect over time is also robust to model misspecification. In survival data characterized with large number of clusters, the test is powerful even if the data is highly heterogenous and/or there is misspecification error.