Abstract
Box proposed an empirical procedure, called the lambda plot, to search for a simplifying transformation in the context of replicated orthogonal experiments. This paper shows that, by appropriately scaling the factor effects, the lambda plot procedure can be extended to unreplicated situations and illustrates its use on four published data sets. The scaling technique proposed can be used also to construct a simple graphical summary of the analysis of multiresponse orthogonal experiments with two-level factors. This is illustrated using data obtained from a pigment production process.
Additional information
Notes on contributors
Y. L. Grize
Dr. Grize is statistician in the Mathematical Applications Department.