Peter H. Westfall (Texas Tech University),

Russell D. Wolfinger, Randall D. Tobias (SAS Inc., USA)

New Bayesian and Frequentist Software Solutions for Multiple Inferences

Our recent book "Multiple Comparisons and Multiple Tests using the SAS(R) System" contains a number of software solutions that are relatively new, and easy to implement using available SAS(R) procedures and macros. These include various Bayesian inferences for functions of mean and variance parameters in random effects models (possibly heteroscedastic), including (i) calculation of posterior probabilities of rankings, (ii) simultaneous Bayesian credible intervals, and (iii) Bayesian decisions using the loss function approach. Bayesian software for calculating posterior probabilities of point null hypotheses in free combination applications also is made available. On the frequentist side, we present software based on Shaffer's "method 2"that utilizes correlations from general models and sets of contrasts, as well as software for determining required sample sizes and estimating directional error rates.