A characterization of the Pareto distribution based on the Fisher information for censored data under non-regularity conditions
- Authors
- Type
- Published Article
- Journal
- Metrika
- Publisher
- Springer Berlin Heidelberg
- Publication Date
- Nov 20, 2018
- Volume
- 82
- Issue
- 4
- Pages
- 429–440
- Identifiers
- DOI: 10.1007/s00184-018-0697-5
- Source
- Springer Nature
- Keywords
- License
- Yellow
Abstract
It is proved that within a proper class of distributions, the Pareto and the shifted exponential distribution are the only distributions with the property of no loss of information due to type-I censoring and random censoring. The equality of the information before and after censoring it is achieved only when the regularity conditions do not hold.
