|
Activity Number:
|
339
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #306694 |
|
Title:
|
Automatic Detection of Outliers Based on the Forward Search
|
|
Author(s):
|
Matilde Bini and Bruno Bertaccini and Franco Polverini*+
|
|
Companies:
|
University of Florence and University of Florence and University of Florence
|
|
Address:
|
Viale Morgagni 59, Florence, 50134, Italy
|
|
Keywords:
|
cut off point ; forward search ; robustness
|
|
Abstract:
|
The article was stimulated by the work of Atkinson and Riani on the estimation of regression models using a robust methodology called "forward search" that seems to work well in the estimation of a variety of models particularly when part of the data are generated by models different from the one we intend to estimate. The methodology detects the presumed outliers and estimates the models without them. The weak point of the procedure is, in our opinion, that the choice of the subset of the data to use for this robust estimation relies on the behavior of some statistics as one adds observations to an initial small set of data. The aim of the paper is to propose some alternative and automatic ways to make this choice that would allows the running of simulations to assess the properties of the estimators and the power of the various tests used in the estimation of the models involved.
|
