Graphic Exploratory Analysis of Variance Illustrated on a Splitting of the Johnson and Tsao Data,
Abstract
Since our main purpose is to illustrate how a combination of graphic display and simple arithmetic can be used to enhance the effectiveness of Daniel's half-normal plots, we shall focus on an analysis of 2x4x7 data set provided by the responses of person IB1 in the 'methodological experiment' presented by (Palmer) Johnson and Tsao (1944) and also analyzed by Palmer Johnson (1949) and by Green and Tukey (1960). THe full data set involves 8 persons. We will return below to a fuller description of the data set and a discussion of how it might be analyzed, and elsewhere to a discussion of how the analysis might be carried further, with special attention to the identification and sterilization of exotic values. We need to make clear our overall prejudices and purposes - a (well-founded) prejudice that most analyses of variance with 3 or more factors are used for exploration rather than confirmation, with a clear description of the apparent behavior being much more important than formal significance tests, and a clear purpose to reach as simple and complete description of the data behavior, in the instance before us, as we know how, taking reasonable account of questions of multiplicity, bu not overemphasizing precise significance. We want a description of the appearance of our data, even though more data might not confirm that appearance.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1987
- Accession Number
- ADA187297
Entities
People
- Eugene G. Johnson
- John W. Tukey
Organizations
- Princeton University