The Relationship between the Number of Factors and Size of an Experiment.
Abstract
If the number of factors in an experiment is increased, does it necessarily follow that the size of the experiment must increase to achieve a satisfactory analysis? In some common situations the answer is No. The present paper discusses a model suggested by medical diagnostic problems in which the answer is Yes: indeed, the increase in size is exponentially fast. The conclusion is drawn that statisticians should be cautions before embarking on the study of data with large numbers of factors because the data may be inadequate for a sensible analysis. The basic, mathematical tool is the Kullback-Leibler number which measures the discrimination between the possibilities. Calculation of these numbers uses interactions, forming a basis for all the effects that might occur. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1981
- Accession Number
- ADA103877
Entities
People
- D. V. Lindley
Organizations
- University of Wisconsin–Madison