Comments and Suggestions on Efficiency Robustness.
Abstract
The introduction sets out some general available procedures to get large-sample efficient estimates of a location parameter when the governing distribution f is well-specified. The next section reviews some attempts to relax the f specification. In the third section is discussed how one chooses from a given repertoire of competing estimators. It is there advocated that their respective estimated standard errors be used to govern the choice and two methods are presented for estimating standard error nonparametrically for this purpose. Some Monte Carlo comparisons are presented in section 4 using sample-sizes of 30, 60, and 120 together with a short-tail and a long-tail f. The possibility of using sample-determined weightings of selected estimators is also briefly explored. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 24, 1970
- Accession Number
- AD0714813
Entities
People
- Paul Switzer
Organizations
- Stanford University