Self-Critical and Robust Procedures for the Analysis of Univariate Complete Data.
Abstract
A statistical sensitivity analysis may be defined and performed in terms of the response of a vector of parameter estimates to variation in the way sample information is processed vis-a vis to tentative underlying model. The mode of information processing is a generalization of likelihood and is indexed on a non-statistical parameter c. The case c = 0 corresponds to maximum likelihood. If the vector of parameter estimates is stable under moderate increase of the index c from 0, the tentative model and the data are internally consistent. A general procedure for the conduct of such sensitivity analyses is given along with several illustrations. For fixed, positive values of the index, one obtains a general robust estimation procedure. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1982
- Accession Number
- ADA119724
Entities
People
- A. S. Paulson
- E. H. Nicklin
- M. A. Presser
Organizations
- Rensselaer Polytechnic Institute