Applications of a Simple Effect Size Estimator for Single-Case Designs

• In Event: 51.029 - Handling Methodological Complexities in Effect Size Calculation and Meta-Analysis of Single-Case Experimental Design Data

Abstract

Multilevel statistical models, both linear and nonlinear, have been developed for analyzing single case experimental design (SCED) data. These models successfully take into account most aspects of the design, and answer research questions of interest to SCED researchers.
Unfortunately, these models are often beyond the understanding of many of the applied researchers for whom they were developed. SCED researchers have typically avoided statistics, and when forced seem to prefer methods involving 2 x 2 tables. I have proposed a simple transformation of the log odds ratio (and its standard error) as a statistic that might appeal to such researchers. In this year’s poster, I will apply that measure in several cases and discuss what the statistic can and cannot do relative to more complicated multilevel models. (For example, multilevel models cannot be applied to scenarios with two cases, and is not well estimated with as few as three or four cases per study. Some overlap indices and measures are undefined when there is no overlap between phases. The log odds ratio can be easily applied in both of these situations).
In the final paper, I will also discuss simple methods of combining the log odds ratios from several cases, in scenarios in which the fixed effect model holds and in other conditions when the fixed effects model does not hold.

Author

• David M. Rindskopf, City College of New York - CUNY