Accessible randomization tests for single-case and small-n experimental designs in AAC research

Abstract

It is argued that valid inferences about the efficacy of treatment effects are as dependent on random assignment procedures in single-case and small-n experimental designs as they are in large-n designs. Traditional single-case and small-n designs can readily be modified to incorporate random assignment of available times or subjects to treatments. True experimental designs generated by random assignment procedures offer the additional advantage that they are susceptible to valid statistical analysis by means of randomization tests based on the random assignment procedure. The importance of using random assignment procedures when evaluating treatment efficacy in single/small-n designs is demonstrated and the randomization approach to statistical testing is described. Finally, examples of single/small-n designs appropriate to AAC research are used to demonstrate the randomization approach. For one of the example designs, a set of Minitab commands (macros) for the appropriate randomization test is provided, a data set is analyzed, and statistical conclusions are drawn. Minitab macros for all of the example designs are available on disk, together with an instruction manual containing worked examples.