Activity Number:
|
357
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 14, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
General Methodology
|
Abstract - #300501 |
Title:
|
Multi-Step Sequential and Accelerated Sequential Methodologies for a Replicable Linear Model
|
Author(s):
|
Greg Cicconetti*+ and Mun Son and Nitis Mukhopadhyay and Yong Ko
|
Affiliation(s):
|
University of Connecticut and University of Vermont and University of Connecticut and University of Vermont
|
Address:
|
PO BOX 179, Storrs, Connecticut, 06268, USA
|
Keywords:
|
ellipsoidal confidence regions ; mulit-step acceration ; conditional dispersion matrix ; fixed-maximum diameter ; largest characteristic root ; oversampling rate
|
Abstract:
|
We consider a sequence of observable p-dimensional iid normal observations with mean AQ and unknown p.d. variance matrix S* where A is a known matrix of rank q and Q is a q-dimensional vector of unknown regression parameters. We propose new multi-step sequential and accelerated sequential procedures for constructing confidence ellipsoids with maximum diameter less than predetermined length 2d. These procedures, which greatly reduce the "oversampling rate," are compared with their predecessor, a two-stage methodology by Chattergee's (1990). Asymptotic properties of five methods are compared. A large simulation study was conducted to investigate performance under moderate sample sizes.
|