Tony Hayter (Georgia Institute of Technology, USA),

Tetsuhisa Miwa (National Institute of Agro-Environmental Institute, Japan),

Wei Liu (University of Southampton, UK)

Combining the advantages of one-sided and two-sided multiple comparison


We consider the multiple comparison problems of making all pairwise comparisons among a set of treatment effects, and comparing a set of treatment effects with a control treatment, through the construction of simultaneous confidence intervals. One-sided procedures have the advantage of indicating the greatest number of significant differences in the direction of interest to the experimenter, whereas two-sided procedures provide both lower and upper bounds on the treatment differences. We present some new procedures which at the same specified confidence level combine the advantages of both the one-sided and the two-sided procedures. The new procedures are illustrated with some examples.