Abstract
In this paper, some nonparametric approaches of density function estimation are developed when censoring indicators are missing at random. A conditional mean score based estimator and a mean score estimator are suggested, respectively. The two estimators are proved to be asymptotically normal and uniformly strongly consistent. The bandwidth selection problem is also discussed. A simulation study is conducted to compare finite-sample behaviors of the proposed estimators.
Original language | English |
---|---|
Pages (from-to) | 835-850 |
Number of pages | 16 |
Journal | Journal of Multivariate Analysis |
Volume | 100 |
Issue number | 5 |
DOIs | |
Publication status | Published - 1 May 2009 |
Externally published | Yes |
Keywords
- 62E20
- 62G05
- Asymptotic normality
- Bandwidth selection
- Mean-squared error
- Missing at random
- primary
- secondary
- Strong consistency
ASJC Scopus subject areas
- Statistics, Probability and Uncertainty
- Numerical Analysis
- Statistics and Probability