Comparison of Several Estimators for the Variance of a Normal Distribution when Outliers May be Present.
Abstract
Although much research has been directed at dealing with outliers, particularly for a N(MU, sigma sq) distribution, little work has been devoted to estimating sigma sq when outliers may be present. This report describes a comparison of some proposed estimators of sigma sq for the case of data which, except for spurious observations, results from a N(MU sigma sq) distribution. All the estimators considered are nonadaptive and can accommodate up to n/2 outliers from a sample of size n. Monte Carlo simulation was used for the comparison of MSE, which was selected as a measure of performance. Interval estimates based on several of the estimators are also considered. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1980
- Accession Number
- ADA087491
Entities
People
- Denise D. Schmoyer
- Dennis E. Smith