Meta-Analyses of Single-Subject Research Designs with R

According to both the National Association of Social Workers (2008) and the Council on Social Work Education (2015), social workers must have the skills to evaluate their practices, including how well clients respond to interventions. Often this is a grassroots effort, and evaluations frequently begin with assessing one client at a time. Single-subject research is one method for social workers to evaluate their practices; however, in many cases, it may be useful to aggregate the findings of multiple, related single-subject studies. This can occur in cases where multiple clients may have similar goals or when similar interventions are used with different clients.

Analysis of single-subject research data may occur on two levels: visually and statistically (Mattaini, 2010; Rubin & Babbie, 2011). Visual analysis is the most commonly used approach (Kazdin, 2011; Orme & Combs-Orme, 2012); however, there are situations in which visual analysis alone may not be sufficient. Statistical analysis may be valuable if observed effects are small, variation is large, effects of the intervention do not appear immediately as phases change, there is no clear trend in the data within phases, or when there may be an issue of autocorrelation, which is nearly impossible to detect through visual inspection alone (Kazdin, 2011; Thyer & Myers, 2011).

The need for statistical analysis in single-subject designs, however, presents a challenge, as analytical methods that are applied to group comparison studies are not appropriate in single-subject research. This is because single-subject research typically compares measured outcome(s) over time under varying conditions, but across the same client unit. Therefore, it is often appropriate to apply different statistical treatments in single-subject designs than in group comparisons that require independence of observations (Kratochwill et al., 2010).

SSDforR is a comprehensive R package designed for analyzing single-subject data both visually and statistically. It has capabilities that may make it suitable for both practice evaluation and advanced, publishable research (Auerbach & Zeitlin, 2014). In its latest iteration, two functions have been added to conduct meta-analyses. The metareg() function allows users to calculate effect sizes and variances across any number of clients while the metaregi() function enables meta-analyses with the inclusion of a moderating variable.

In the first part of this workshop, attendees will be shown how to freely obtain and utilize the basics of the package. This will include a demonstration of how to import single-subject data into SSDforR, conduct analyses based on characteristics of the data, and interpret findings that can guide practitioner decision-making. In the second part of the workshop, the presenters will focus on the utilization of the metareg() and metaregi() functions. A discussion of how these could improve published research will ensue with examples from previously published research studies.

Participants will be provided with resources that will enable them to use all functions in SSDforR after the workshop including presentation materials, instructions for downloading R, SSDforR, scripts used during the presentation, and sample data.