Premium
Reducing the Impact of Missing Values in Factorial Experiments Arranged in Blocks
Author(s) -
Godolphin J. D.
Publication year - 2006
Publication title -
quality and reliability engineering international
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.913
H-Index - 62
eISSN - 1099-1638
pISSN - 0748-8017
DOI - 10.1002/qre.804
Subject(s) - factorial experiment , fractional factorial design , factorial , missing data , design of experiments , hazard , computer science , statistics , process (computing) , product (mathematics) , vulnerability (computing) , mathematics , reliability engineering , engineering , mathematical analysis , chemistry , geometry , computer security , organic chemistry , operating system
The loss of one or more observations during a designed experiment is a hazard well known to statistical practitioners, including practising engineers involved in product design and development. When observations are missing the eventual experimental design is necessarily altered from the design that was selected originally. This can introduce unwelcome difficulties, even though the available software may allow for missing values in the analysis of the surviving data. In particular, if missing values are incurred in a replicated factorial experiment arranged in blocks, the eventual design may be disconnected with respect to some factors under investigation and the experiment may be seriously damaged. It is suggested that the vulnerability of the original design should be examined at the planning stage of the experiment so that suitable action can be taken before any experimentation is done. These points are illustrated by considering an experiment consisting of three replicates of a full 2 3 factorial arranged in rows and columns and it is shown how to recognize the vulnerability of designs before the experimental process has taken place. Copyright © 2006 John Wiley & Sons, Ltd.