Randomization tests have been suggested as a method for analyzing the data from single-case designs. This paper explores three concerns which have arisen: the nonresponsive nature of randomization tests, the appropriateness of specific test statistics and the power of these tests. An example is given to illustrate how partial control can be given to the researcher so that a design can be responsive and incorporate random assignment. A general test statistic is given that can be used when the researcher is unsure of the specific nature of the treatment effect and thus uncomfortable with the specific test statistics typically employed. Finally a method of combining types of randomization is given to provide a way of increasing the number of assignments and the power o the randomization test. (Author) Reproductions supplied by EDRS are the best that can be made from the original document.

Eight high school students who were peer tutors in a classroom for students with severe disabilities were matched in pairs and then randomly assigned to interact with and serve as a partner for a classmate with severe disabilities. One participant in the pair received social interaction training, and one participant did not. Statistical analysis indicated that the social interaction training increased the frequency of initiations directed from the nondisabled students toward their partners with severe disabilities. Additionally, there was an increase in the proportion of interactions that were social in nature, with a resulting decrease in the frequency of task-related interactions. There was also a significant increase in targeted social behaviors of the participants with severe disabilities associated with the increase in social interactions with their nondisabled partners. (Contains 40 references.) (Author/JDD) Reproductions supplied by EDRS are the best that can be made from the original document.