Performance of Density Estimators in Additive Measurement Error Models Based on Right Censored Data
View/ Open
Date
2022-05-09Author
Khakurel, Hrishabh
0000-0001-5328-8034
Metadata
Show full item recordAbstract
**Please note that the full text is embargoed until 5/6/2024** ABSTRACT: In the deconvolution problem for right censored data, one is interested in estimating the density of a contaminated variable X when X satisfies Z= X+ E, where E is a measurement error with a known distribution, and the observable variable Z is right-censored. Zhu, Sun, Khakurel, and Wang (2021) applied the Inverse Probability of Censoring Weighted Average method and derived the estimators of the unknown density of X. In this study, we evaluate the performance of the density estimators both in theory and in simulation. We derive the theoretical upper bounds for Mean Squared Error (MSE) of the estimator and its derivatives, accounting for the tail behavior of the error distribution. Our simulation studies focus on: (a) the problem of estimating the unknown censoring distribution, (b) methods of selecting the optimal bandwidth, and (c) the effects of the kernel and error distributions on the density estimators. Our simulations show that the estimators perform reasonably well when sample sizes are relatively large.