Design and Analysis of Experiments for Statistical Selection, Screening, and Multiple Comparisons

A practical guide to selection, screening, and multiplecomparisons This book addresses experimenters who have knowledge of classicalexperimental design methodology and expands their repertoire beyondhypothesis testing by providing statistical methods appropriate forselection, screening, and multiple comparisons. It concentrates onthree types of procedures: selection procedures that use the"indifference-zone" approach, screening procedures using the"subset" approach, and multiple comparison procedures involvingnormal means. This is the first book, specifically designed forpractitioners, to bring into focus many developments in the fieldpreviously covered only in university courses. It also presents newresults on the comparison of procedures that have been obtainedspecifically for this volume. This self-contained volume describes methods for designingexperiments when the scientific objective is selection of besttreatments, screening a set of treatments, and multiple comparisonsamong treatment means. The book emphasizes procedures appropriatein a variety of practical settings including those that requireblocking and randomization restriction. It compares the relativemerits of procedures when several different methods can be used inthe same circumstances. Providing practical guidance for experimenters in agriculture,engineering, medicine, and other empirical sciences, this book mayalso be used for a one-semester graduate course in selectionmethodology or to augment traditional courses in experimentaldesign. Design and Analysis of Experiments for Statistical Selection,Screening, and Multiple Comparisons:* Shows how selection and screening can be applied to data thatfollow one of three important probability models--normaldistribution, binomial distribution, and the multinomialdistribution models* Provides an extensive comparison of procedures, allowingexperimenters to choose among competitors when several differentprocedures are feasible for a given application* Gives an extensive set of tables of constants necessary toimplement the procedures* Supplements the tables of constants with listings of FORTRANprograms so that experimenters are not limited to those valuescovered by the tables* Focuses on frequent formulations, while also providing referencesto Bayesian and other alternative developments in the Chapter Notes