Review of Functional Forms in Intervention Phases of Multiple-Baseline Designs

• In Event: Handling Methodological Issues in the Analysis of Single-Subject Experimental Design Data

Abstract

There has been a surge of recent methodological research focused on analyzing single-subject experimental design (SSED) data using parametric statistics (Beretvas, Hembry, Van den Noortgate, & Ferron, 2013; Hedges, Pustejovsky, & Shadish, in press; Shadish, Kyse, & Rindskopf, in press; Ugille et al., 2012; Van den Noortgate, 2008) with most studies focused on the multiple-baseline design (MBD). Depending on the behavior being measured as the outcome and on its functional relationship with the intervention (among other study and sample characteristics), the researcher might have very different hypotheses about what the treatment affects. A treatment can affect several different facets of an outcome’s trajectory in SSED data. For example, the treatment might change the level of an outcome; it might change the linear growth or even introduce linear growth after a flat, stable baseline. The treatment could introduce a nonlinear trajectory that culminates in an upper or lower asymptote being reached. Any of these intervention effects might begin after a delay.
Due to the complexities being addressed in methodological research on SSED data analysis, the majority of the methodological research has focused on models or effect sizes that are based on an assumption that the treatment changes the level and/or linear slope of the outcome. A couple of studies have tested or demonstrated models and effects for scenarios with nonlinear treatment phase trajectories (e.g., Beretvas et al., 2013; Nagler, Rindskopf, and Shadish, 2008). However, it is unclear what constitute typical treatment phase trajectory patterns. As methodological research on SSED analyses continues to increase, it would be helpful to understand the kind of treatment phase trajectories that methodologists should consider and address in their research.
The current study has been designed to survey and synthesize recent applied MBD research to provide information about treatment phase trajectory forms. This information can then be used by methodologists to help them better understand and design authentic real data scenarios in their simulation studies. The design of this review matches closely that done by Shadish and Sullivan (2011) in that the most recent full year (2012) of applied MBD studies will be reviewed. The associated text of the MBD studies will be surveyed to gather information how the authors describe expected intervention phase trajectory patterns (including form –flat, linear, nonlinear, increase, decrease, etc. – and whether a treatment should take a while to have its effect, etc.). In addition, intervention phase data will be gathered using the Data Thief software and entered into a database. The best description of the data’s functional form (flat, linear slope, quadratic, logistic, etc.) will be estimated across participants within each study. The results will be summarized and the implications discussed.
Study selection will be completed by the end of August, 2013. Data will be gathered and entered into the database by the end of December and summarized in time for a final paper that will be submitted to the Discussant by March, 2014.

Authors

• Rommel Lazo Bunuan, The University of Texas - Austin

• Susan Natasha Beretvas, The University of Texas - Austin