The Use of Influence Functions for Outlier Detection and Data Editing.
Abstract
Hampel's influence function (J. Amer. Statist. Assoc. 69, pp. 383-393) has been used in recent years for evaluating robust estimators, detecting outliers, computing asymptotic variances for estimators, and for hypothesis testing. In this report, the use of influence functions for various parameters is proposed, not only as a tool for outlier detection, but also as a method for replacing outliers and/or missing observations. This approach is illustrated on the estimation for the variance of a distribution and it is pointed out how the method can be applied in simple linear regression problems. This approach can be extended to more complex regression type applications such as the estimation of the orbit parameters of a satellite.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 15, 1982
- Accession Number
- ADA120561
Entities
People
- M. R. Chernick
- V. K. Murthy
Organizations
- The Aerospace Corporation