On the robustness of right and middle censoring schemes in parametric survival models

This article considers the exponential and Weibull distributions for studying the robustness of their parameters estimates in the case of right and middle types of censoring. As might be expected, the simulation studies indicate that the mean square error (MSE) decreases with larger sample sizes, increases with the contamination level and censoring proportion. But somewhat surprisingly, our study shows that, regardless the parametric model and for given values of the other factors, estimates for middle censoring are more robust compared to those for right censoring.