Monograph Series #2
Multiple Comparisons by Multiple Linear Regressions
By John D. Williams, The University of North Dakota
Abstract: Several of the more common multiple comparison techniques are explored in a regression approach. Dunnett's test for comparing several groups to a single group, Tukey's(a) honestly significant different test, Newman-Keul's, Tukey's(b) and Duncan's tests are considered. Complex comparisons (contrasts) are shown through Dunn's and Scheffd's test and through orthogonal comparisons.
Orthogonal polynomials are also shown for testing for trend. A method for finding a maximized Scheff6 contrast such that the contrast will yield the same R value as the original full model is also included.
The intent of the present monograph is to more fully explore the use
of alternate methodologies to the usual multiple F tests when more than
one restriction is placed on a full model.