Using SAS for statistical modeling: MONTE CARLO simulations

The development of new statistical formulae generally requires various mathematical derivations to clearly prove the accuracy and usefulness of the new formulae. However, an additional verification of the new formulae is also becoming standard and that is Monte Carlo simulation. Monte Carlo simulations provide a procedure for comparing new formulae versus current formulae and a forum to assess their relative effectiveness. The SAS(R) statistical computing software is very useful in this process and the current paper provides a model of how this can be completed. A series of new formulae for computing statistical power are generated and compared with a current procedure using Monte Carlo simulation. The procedures used in SAS for this process are PROC IML, PROC APPEND, PROC MEANS, and PROC ANOVA. The programs and statistical analyses are appropriate for all levels of SAS users and in particular those interested in statistical modeling.