| Summary: | The writers describe an SAS computer program, based on Scheffe's well-known multiple comparison procedure, for computing comparisons among k independent binomial samples. As in Scheffe's “data snooping” analytic technique, the comparisons can be planned or post hoc and can also be simple or complex. Although the comparisons do not have to be orthogonal, none will be statistically significant unless the omnibus statistic, which is chi-square, is statistically significant. In general, this SAS program can be utilized when the design is not balanced. Details regarding the program's availability are provided.
|
|---|