Analysis of Unreplicated Factorials Allowing for Possibly Faulty Observations.
Abstract
Normal probability plotting of the orthogonal contrasts (Daniel (1959)) is a useful diagnostic as well as inferential technique for analyzing unreplicated factorial experiments. In particular, the presence of one of more faulty observations is marked by a characteristic defect of the normal plot. In previous work (Box and Myer (1985)), the authors introduced a more formal method of identifying active effects as a supplement to Daniel's graphical analysis. This work is extended here to allow for the possibility of faulty observations. Keywords: Unreplicated factorial, Normal probability plot, Faulty observation, Detection, Accommodation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1985
- Accession Number
- ADA154813
Entities
People
- George E. P. Box
- R. D. Meyer
Organizations
- University of Wisconsin–Madison