A new nonparametric bivariate survival function estimator under random right censoring
MetadataShow full item record
This thesis examines the efficiency and accuracy of a new nonparametric bivariate survival estimator under right censoring. The estimator and its doubly modified version were simulated in a number of design settings to investigate their finite sample behaviors. Both S-Plus and Fortran 90 codes are employed in the simulation to speed up the running of program. The bias, standard deviation and mean squared error are investigated to study the efficiency and accuracy of estimators with contour and wireframe plots. The results show that, overall, the new bivariate survival estimator and its modified version work fairly well. These new estimators overcome a number of drawbacks in several earlier nonparametric estimators in terms of their finite sample properties, such as monotonicity, simple computation and reduction to the empirical distribution in the uncensored case.