In Chapters 4 through 9 we have presented power and sample size tables that can be addressed directly with no preliminary adjustments to the effect size or N. Unfortunately the complexity of some research designs introduces scenarios in which this convenience is not always feasible, hence in this chapter we present a number of tables and algorithms by which both the ES and N can be adjusted to permit the reader to adapt the tables presented in previous chapters. Unfortunately, our guidelines in this chapter will not be quite as explicit as those provided to this point – because the techniques for modeling the power of complex designs are not well developed and there is little consensus regarding the appropriateness for those that do exist. Our advice for employing the modeling procedures presented in this chapter, therefore, is to approach the task from the perspective that any accruing results will truly be estimates (as all power/sample size estimates are). It is also a good practice, when communicating these results, to present both a conservative estimate (based upon the procedures employing relatively fewer assumptions presented in previous chapters) along with these modeled estimates.

To this point, then, we have provided tables applicable to designs involving a continuous dependent variable coupled with (a) a single grouping variable involving from two to five groups (with and without repeated measures), (b) the interaction between two grouping variables when no more than one of these variables involves a repeated measure, and (c) the addition of one or more covariates to all of these designs which do not involve repeated measures.