G. Vallejo, P. Livacic-Rojas and N. Conejo
At present the randomization tests are considered as suitable statistical methods for evaluating the data from multiple baseline AB designs. Particularly, when the researchers, in addition to determining the intervention points at random, also randomly selects the order in which the person, behaviors, or situations are allocated to the treatment programs. However, in order to conduct a valid randomization test it is required to randomize some aspect of the design, which unfortunately usually is not frequent in the applied clinical investigation. This is a situation in which a randomized test is inadequate and some alternative procedure may be preferable in order to draw valid statistical inferences about treatment effects. In this article a pooled interrupted time series method is presented for the analysis of data based on these designs, and an example is given to illustrate the corresponding technique.
AB designs, Pooled time series