Testing Whether one Regression Function is Everywhere Larger than Another.
Abstract
The problem of testing whether one regression function is larger than another on a specified compact set R is considered. The regression functions must be linear functions of the parameters but need not be linear functions of the independent variables. The proposed test statistic is compared to a standard t percentile. The test has an exactly specified size in typical situations. Properties of the power function of the test are investigated. The related question of comparing a regression function to a specified function is also considered. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1982
- Accession Number
- ADA112212
Entities
People
- Roger L. Berger
Organizations
- Florida State University