EDF Statistics for Goodness-of-Fit: Part 1
Abstract
The paper deals with EDF statistics, a class of goodness-of-fit statistics. They are based on a comparison of F(x) with the empirical distribution function (F sub n)(x). Recent work has now made it possible to use these statistics very easily when F(x) is continuous and completely specified and also for two very important practical situations - when the distribution tested is normal, or exponential, with parameters to be estimated. Power studies suggest they should be brought into wider use. The report concentrates on a practical guide to the use of EDF statistics; specifically, those usually called D(+), D(-), D, W(2), V, U(2), A.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 03, 1972
- Accession Number
- AD0737653
Entities
People
- M. A. Stephens
Organizations
- Stanford University