Influence analysis in truncated distributions |
| |
Authors: | I. Barranco-Chamorro,J.L. Moreno-Rebollo,J.M. Muñ oz-Pichardo |
| |
Affiliation: | a Dpt. Estadística e I.O., Fac. de Matemáticas, Univ. de Sevilla, C/Tarfia, s.n., 4012 Seville, Spain b Centro Andaluz de Prospectiva, C/Tarfia, s.n., 4012 Seville, Spain |
| |
Abstract: | Conditional bias and asymptotic mean sensitivity curve (AMSC) are useful measures to assess the possible effect of an observation on an estimator when sampling from a parametric model. In this paper we obtain expressions for these measures in truncated distributions and study their theoretical properties. Specific results are given for the UMVUE of a parametric function. We note that the AMSC for the UMVUE in truncated distributions verifies some of the most relevant properties we got in a previous paper for the AMSC of UMVUE in the NEF-QVF case, main differences are also established. As for the conditional bias, since it is a finite sample measure, we include some practical examples to illustrate its behaviour when the sample size increases. |
| |
Keywords: | AMSC Conditional bias Influence function Truncated models UMVUE |
本文献已被 ScienceDirect 等数据库收录! |
|