Abstract
The nature and uses of the split-plot design are discussed. An initial section reviews models and least squares to establish a point of view. The major section presents the development of a simple split-plot design by a process of “evolution,” beginning with a completely randomized model and progressing through a randomized blocks design to a split plot, using the same set of runs throughout. It is shown that the three designs differ only in the randomization procedures used. This emphasizes the importance of proper randomization and use of the correct analysis layout and indicates how readily an experimenter may inadvertently produce a split-plot experiment which is unrecognized as such, resulting in the use of an incorrect analysis and exaggerated or wrong conclusions. A final section gives an example of an actual split-plot experiment and shows the analysis of the data.
Additional information
Notes on contributors
W. M. Wooding
Mr. Wooding, Director of Technical Services at Carter Products Research, Carter-Wallace, Inc., Cranbury, N.J., is a Senior Member of ASQC.