|
Activity Number:
|
169
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Education
|
| Abstract - #305530 |
|
Title:
|
Misconceptions and Properties of Friedman's Test
|
|
Author(s):
|
Philip Turk*+ and Roy St. Laurent
|
|
Companies:
|
West Virginia University and Northern Arizona University
|
|
Address:
|
Department of Statistics, Morgantown, WV, 26506,
|
|
Keywords:
|
Friedman's Test ; Nonparametric Test ; Analysis of Variance ; Skewness ; Heteroscedasticity ; Asymmetry
|
|
Abstract:
|
Friedman's test is a widely used rank-based alternative to the analysis of variance F-test for identifying differences in treatment means or medians in a randomized complete block design. Many introductory texts that discuss Friedman's procedure provide incomplete or misleading information about when it may be appropriately applied. In this paper, we discuss the assumptions needed for Friedman's test and common misconceptions. We show via Monte Carlo simulation that when the variance or skew of the treatment distributions differ, application of Friedman's test to detect differences in treatment location (mean or median) can lead to situations where the actual level of the test is substantially larger than the chosen nominal alpha. We explore the connection between a measure of asymmetry of the treatment rank distributions and the observed level of the test.
|
