Robust Data Analysis Based on Characteristic Functions.
Abstract
Simultaneous estimators of location and scale parameters of a Gaussian population are obtained from an analogue to weighted least squares by placing the sample characteristic function and the population characteristic function in the role of observed and expected quantities. The advantage to such a scheme is that the error and structural model play a direct role in the analysis. The analogue to the usual sum of squares, the sum of moduli squared, then depends on a nonstochastic nusiance variable u. Removal of the nusiance variable is effected by integration over u. A simulation study provides some small sample properties of the location and scale estimators. An example showing the potential for complementing response surface methodology is provided. Keywords: tables(data); distribution functions; robustness; scale parameters; regression atuoregression.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1986
- Accession Number
- ADA178544
Entities
People
- A. S. Paulson