|Current Issues in Behavioral Measurement|
|Monday, May 28, 2012|
|3:30 PM–4:50 PM |
|Area: DDA/EDC; Domain: Applied Research|
|Chair: Jeffrey H. Tiger (University of Wisconsin, Milwaukee)|
|Discussant: Gina Green (Association of Professional Behavior Analysts)|
|CE Instructor: Jeffrey H. Tiger, Ph.D.|
The refinement of techniques to produce accurate and reliable data represent an important applied endeavor. The presentations included in this symposium are directed toward ensuring the highest levels of accuracy in measuring the frequency, duration, and magnitude of behavior.
|Keyword(s): accuracy, calibration, measurement, response magnitude|
The Application of Time and Motion Study Methods to Determine the Representativeness of Duration Measures in Observation Samples
|REBECCA SHARP (University of Auckland), Oliver C. Mudford (University of Auckland), Douglas Elliffe (University of Auckland)|
Behavior analysts have sought to develop effective methods of producing observation samples that reflect overall dimensions of behaviour across the whole time of interest. Little research has been conducted in behavior analysis in how to obtain representative samples, although it has been demonstrated that as sample length decreases, error increases, with greater error in the sampling of low-duration behaviours (Mudford, Beale, & Singh, 1990). Work sampling (a method in time and motion study) is used to determine how people spend their time in work settings in an efficient way and thus may assist behavior analysts to empirically determine the length of samples required to produce representative data. In the current study, full time-of-interest (a full school week in a special school) direct observations of several behavioral and environmental categories in children with special needs were conducted. Formulae used in work sampling were then applied to determine the length, number, and time of samples required for the data to be representative. Staff reports and preliminary observations were used in the work sampling formulae to calculate the initial number of observations to be conducted for a representative sample of each category. Findings and the utility of work sampling methods in behavior analysis will be discussed.
The Utility of Interobserver Agreement and Calibration in Assessing Quality of Behavioral Data
|KATRINA J. PHILLIPS (University of Auckland), Oliver C. Mudford (University of Auckland), Douglas Elliffe (University of Auckland)|
Applied behaviour analysis typically uses interobserver agreement (IOA) algorithms as a measure of our confidence in the quality of the data. Other natural sciences use calibration to assess the quality of the data produced by their measurement systems. The current study conducted calibration analysis and three IOA assessments (block-by-block, exact agreement, and time-window analysis) for novice recorders using different recording methods (laptop computer, handheld touch screen computer, and events within intervals using pen and paper). The presentation will discuss the strengths and limitations of IOA and calibration based on the data from 15 observers. Recommendations for selecting and evaluation measurement systems in applied behaviour analysis will be made.
Objective Measure of Motor Movements
|ANDREA R. REAVIS (Marcus Autism Center), Nils Y. Hammerla (Newcastle University), Nathan Call (Marcus Autism Center), Thomas Ploetz (Georgia Institute of Technology), Agata Rozga (Georgia Institute of Technology)|
Data on problem behavior (e.g., aggression, self-injury, stereotypy) are typically collected by direct observation using predetermined operational definitions. These definitions are subjective, somewhat arbitrary, and only those behaviors that are directly observable can be recorded. Furthermore, objectively assessing the magnitude (i.e., strength, force, or intensity) of a problem behavior by human observation alone poses significant challenges. The purpose of the current study is to explore the extent to which acceleration data collected from small sensors worn on the wrists and ankles can be used to automatically detect and quantify specific types of problematic behaviors. Data has been collected from an individual who simulated high intensity aggressive, disruptive, and self-injurious behavior while wearing the sensors, as well as with3 individuals who attend a day treatment program for individuals with disabilities who engage in problem behavior. Preliminary analyses of some of the data showed that, when sensor data were compared to data collected by a human observer, the sensors identified instances of problem behavior with 97% sensitivity (true positives) and 84% specificity (true negatives).