|A Crisis in Visual Analysis: Examining Current Analytical Practices in ABAB and Multielement Designs|
|Sunday, May 28, 2017|
|10:00 AM–10:50 AM |
|Convention Center Four Seasons Ballroom 2/3|
|Area: PRA/PCH; Domain: Applied Research|
|Chair: Douglas E. Kostewicz (University of Pittsburgh)|
|Discussant: William L. Heward (The Ohio State University)|
|CE Instructor: Douglas E. Kostewicz, Ph.D.|
Applied and basic researchers in behavior analysis depend on the visual analysis of graphic data. Visual analysis occurs across two phases: a within condition analysis and a between condition analysis. The roots for visual analysis began at the inception of behavior analysis. However, a growing concern within single-case design exists. Namely, the use of supplemental statistics for graphed time series data. Many criticisms lobbied against single case design stem from a lack of universal decision rules and unreliability across raters illustrate the limitations of visual analysis. Critics suggest adding statistical or quantitative analyses to visual analysis provides objectivity, increased confidence of the results, and enhances the strength of the outcome. The present symposium presents two empirical reviews. First, a survey asked two questions: (1) how many visual analytic tactics did the experimenters employ in a within condition analysis and (2) did the experimenters report quantitative or qualitative analyses. The second asked if after recharting functional analysis data on a functional analysis celeration chart, could a numerical value be established to quantify function.
|Instruction Level: Basic|
|Keyword(s): Graphing, Visual Analysis|
Assessing Within Condition: Graphical Analysis Practices for ABAB Designs
|DOUGLAS E. KOSTEWICZ (University of Pittsburgh), Richard M. Kubina (Penn State)|
The analysis of graphic data in single case designs serves as the primary means for judging the significance of results. Single case design books provide guidelines for analyzing data within conditions. The four noted areas targeted for analysis include level, trend, variability, and the number of data points within the condition. The present survey examined 50 articles using ABAB designs within behavioral journals. The survey asked two questions. First, how many visual analytic tactics did the experimenters employ per condition. And second, the quantitative and qualitative analyses reported per condition. The results found a majority of articles used very few within condition analytic tactics. When used, researchers tended to rely on level rather than trend, variability, or number of data points per condition. The experimenters also relied on even fewer qualitative rather than quantitative descriptions of level, trend, and variability. The discussion covers adherence to graphical analysis guidelines and the subjective nature of qualitative descriptors.
Quantifying Function Within Functional Analyses Using Multielement Designs
|RICHARD M. KUBINA (Penn State), Douglas E. Kostewicz (University of Pittsburgh), Sal Ruiz (The Pennsylvania State University)|
Functional Analysis has provided many benefits to many individuals and the field of Applied Behavior Analysis as a whole. Most notably, functional analyses have become an effective assessment tool to discover the environmental variables that maintain challenging behavior. Once determined, interventions can then be created to incorporate the noted variables. Although implementing and analyzing functional analyses has many benefits, many areas for improvement remain. For instance, determining function relies on visually comparing different levels, variability, and the direction and degree of trend. All of the previously mentioned tactics rest on each reviewers subjective impression of the data. Quantifying the function would serve as means to more objectively determine function. The present study examined a series of multi element designs used in functional analyses for the Journal of Applied Behavior Analysis. The data were recharted on a functional analysis celeration chart. The data reveal a range of values indicating the function behavior as revealed through quantification.