This book is a practical guide to help researchers draw valid causal inferences from small-scale clinical intervention studies. It should be of interest to teachers of, and students in, courses with an experimental clinical component, as well as clinical researchers.
Inferential statistics used in the analysis of group data are frequently invalid for use with data from single-case experimental designs. Even non-parametric rank tests provide, at best, approximate solutions for only some single-case (and small-n ) designs. Randomization (Exact) tests, on the other hand, can provide valid statistical analyses for all designs that incorporate a random procedure for assigning treatments to subjects or observation periods, including single-case designs. These Randomization tests require large numbers of data rearrangements and have been seldom used, partly because desktop computers have only recently become powerful enough to complete the analyses in a reasonable time. Now that the necessary computational power is available, they continue to be under-used because they receive scant attention in standard statistical texts for behavioral researchers and because available programs for running the analyses are relatively inaccessible to researchers with limited statistical or computing interest.
This book is first and foremost a practical guide, although it also presents the theoretical basis for Randomization tests. Its most important aim is to make these tests accessible to researchers for a wide range of designs. It does this by providing programs on CD-ROM that allow users to run analyses of their data within a standard package (Minitab, Excel, or SPSS) with which they are already familiar. No statistical or computing expertise is required to use these programs. This is the "new stats" for single-case and small-n intervention studies, and anyone interested in this research approach will benefit.
Inferential statistics used in the analysis of group data are frequently invalid for use with data from single-case experimental designs. Even non-parametric rank tests provide, at best, approximate solutions for only some single-case (and small-n ) designs. Randomization (Exact) tests, on the other hand, can provide valid statistical analyses for all designs that incorporate a random procedure for assigning treatments to subjects or observation periods, including single-case designs. These Randomization tests require large numbers of data rearrangements and have been seldom used, partly because desktop computers have only recently become powerful enough to complete the analyses in a reasonable time. Now that the necessary computational power is available, they continue to be under-used because they receive scant attention in standard statistical texts for behavioral researchers and because available programs for running the analyses are relatively inaccessible to researchers with limited statistical or computing interest.
This book is first and foremost a practical guide, although it also presents the theoretical basis for Randomization tests. Its most important aim is to make these tests accessible to researchers for a wide range of designs. It does this by providing programs on CD-ROM that allow users to run analyses of their data within a standard package (Minitab, Excel, or SPSS) with which they are already familiar. No statistical or computing expertise is required to use these programs. This is the "new stats" for single-case and small-n intervention studies, and anyone interested in this research approach will benefit.
