Statistical Alternatives for Single-Case Designs
Bernard S. Gorman
David B. Allison
Obesity Research Center, St. Luke's/Roosevelt Hospital Center Columbia University Collegeof Physicians and Surgeons
This chapter discusses statistical approaches for analyzing single-case research designs.1 These designs and their corresponding statistical analyses span a continuum that ranges from experimental designs to correlational designs. Figures 6.1a and 6.1b display a typology of analyses. By experimental designs we mean designs in which researchers manipulate one or more independent variables and in which potentially confounding variables are controlled by randomly assigning treatment conditions to subjects. In quasi-experimental designs, the independent variables are manipulated by natural or historical events. Consequently, there is little direct control of unwanted variation, and treatments are not assigned at random. As one departs from the strict requirements of experimental designs, it becomes more difficult to draw precise conclusions about causal relations between manipulations and outcomes.
We can also draw a distinction within the group of experimental designs between those designs that we call "time-series experimental designs" and those that we call "nontime-series experimental designs." Time-series experimental designs measure behavior over continuous____________________