|
| Real-time Data Collection: Impact of Context and Data Collection Methods on the Measurement and Analysis of Behavior |
| Tuesday, June 1, 2004 |
| 10:30 AM–11:50 AM |
| Constitution A |
| Area: DDA/EDC; Domain: Applied Research |
| Chair: A. Katherine Matzen (University of Florida) |
| Discussant: Benjamin W. Smith (University of Texas at Austin) |
| Abstract: . |
| |
| How Much is Enough? An Examination of Behavioral Sampling |
| MAUREEN CONROY (University of Florida), Jennifer M. Asmus (University of Florida), Crystal Ladwig (University of Florida), William H. Brown (University of South Carolina at Columbia) |
| Abstract: Precise and accurate measurement of behavior is a critical aspect of ABA. Collecting a representative sample of behaviors during is essential to provide an accurate description of behaviors (Moore, 1998). A limitation of many current observational procedures is the potential for inaccurate or insufficient sampling of behavior (Freeman, Anderson, & Scotti, 2000). Currently, there is no consensus in the literature regarding the length of observation time needed to obtain a representative sample of behavior. As a result, many decisions regarding behavioral observations are based on the judgment of researchers (Foster & Cone, 1986). The purpose of this study was to examine the differences in rates of behavior across 9 hours of direct observation. The stability of three behaviors (i.e., stereotypy, disruption, and engagement) was compared across 5 different observation lengths (i.e., 1, 3, 5, 7, and 9 hours). Twelve behavior samples were collected across seven settings. Real time data was collected using laptop computers and analyzed using MOOSES (Tapp, 2002). IOA was obtained across 25% of sessions and averaged 85%. Preliminary results indicate that although rate of behavior appears to influence the stability of the behavior across lengths of observation, behaviors remained relatively stable after three hours of observation. |
| |
| Contextual Influences on Sequential Relations in the Study of Classroom Interactions |
| JOSEPH H. WEHBY (Peabody College of Vanderbilt University) |
| Abstract: Over the last several years, there has been increasing use of sequential analysis techniques for identifying relevant antecedents and consequents of both teacher and student behavior in classrooms (e.g., Sutherland, Wehby, & Yoder, 2002; Wehby, Symons, Shores, 1995). Unfortunately, less information is available on the impact of contextual variables, such as type of activity or group size, on these significant relations. This limitation may result in inaccurate assumptions regarding the relation between specific types of teacher and student behavior within different classroom situations. The purpose of this study was to identify differences in sequential relations, as measured by Yules Q, between teachers’ instructional behavior and acts of student aggression and disruption. Observational data from 20 classrooms for students with emotional and behavioral disorders were analyzed using the Multiple Option Observation System (MOOSES: Tapp & Wehby, 2002). Agreement checks were collected on 20% of the observation sessions with an average kappa coefficient greater that .80. Results showed significant differences in Yules Q when significant teacher-student relations were compared under different contextual conditions. Implications for future research in this area will be discussed. |
| |
| Observational Research Methodology, Sequential Analysis |
| JOHN D HOCH (University of Minnesota), Jennifer J. McComas (University of Minnesota), Frank J. Symons (University of Minnesota) |
| Abstract: The results of sequential analyses based on real-time data collection may not be comparable across research groups because coding rules and methods for performing sequential analyses appear to be idiosyncratic. This study examines how various methods of data collection and sequential analysis may influence the obtained results. Using an extant database comprised of direct observational data collected in real time from both live and archived videotape and from a variety of real world contexts, the data were (a) evaluated for reliability using the a/a+d formula, a 5-second window between observers, as well as Cohen’s Kappa and (b) reanalyzed to determine the effects of manipulating time- versus event-based lags. Additionally, for several participants, the proportional matching equation was used to evaluate the results of the different sequential analysis options. The effects of coding behaviors such as self injury as single frequencies for each occurrence or durations of bouts will also be compared within each subject. Interobserver agreement ranged from .74 to .99 on at least 25% of sessions for all participant data used. Cohen’s kappas ranged from .67 to 1.0 on all duration codes. The implications for research and clinical applications will be discussed. |
|
| |