Outlier Resistant Model Robust Regression
Parametric regression fitting (such as OLS) to a data setrequires specification of an underlying model. If thespecified model is different from the true model, then theparametric fit suffers to a degree that varies with the extentof model misspecification. Mays and Birch (1996)addressed this problem in the one regressor variable casewith a method known as Model Robust Regression(MRR), which is a weighted average of independentparametric and nonparametric fits to the data. This paperwas based on the underlying assumption of "well-behaved"(Normal) data. The method seeks to take advantage of thebeneficial aspects of the both techniques: the parametric,which makes use of the prior knowledge of the researchervia a specified model, and the nonparametric, which is notrestricted by a (possibly misspecified) underlying model.The method introduced here (termed Outlier ResistantModel Robust Regression (ORMRR)) addresses thesituation that arises when one cannot assume well-behaveddata that vary according to a Normal distribution.ORMRR is a blend of a robust parametric fit, such asM-estimation, with a robust nonparametric fit, such asLoess. Some properties of the method will be discussed aswell as illustrated with several examples.
Year of publication: |
1997-04-14
|
---|---|
Authors: | Assaid, Christopher Ashley |
Other Persons: | Jeff Birch (contributor) |
Publisher: |
VT |
Subject: | Statistics |
Saved in:
freely available
Saved in favorites
Similar items by subject
-
Fleming, Michael C., (1995)
-
Wilks, Samuel S., (1950)
-
Rossija i strany mira : statističeskij sbornik
(1994)
- More ...