|
Activity Number:
|
446
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #300923 |
|
Title:
|
Estimation and Testing of Linear Treatment Effects from Matched Pair Data
|
|
Author(s):
|
Cornelis J. Potgieter*+ and Fred Lombard
|
|
Companies:
|
University of Johannesburg and University of Johannesburg
|
|
Address:
|
PO Box 524, Johannesburg, International, 2006, South Africa
|
|
Keywords:
|
Matched pairs ; Quantile comparison function ; Location and scale parameters
|
|
Abstract:
|
In the analysis of matched pairs it is standard practice to reduce the data to pairwise differences and to identify the mean difference with the treatment effect. In this talk a more general measure of treatment effect, the quantile comparison function (QCF) which is essentially a QQ-plot is considered in the context of paired samples (Lombard (2005), Technometrics). Specifically, we consider a straight line QCF which is equivalent to a treatment effect consisting of a change of location and/or scale. Permutation tests for the straight-line hypothesis are considered and their powers compared via Monte Carlo simulation. Assuming that the QCF is a straight line, we consider joint estimation of the intercept and slope using sample moments and quantiles and minimum distance methods. Application of the procedures is illustrated in the analysis of parallel coal samples from a common source.
|
