Temporal distributions of problem behavior based on scatter plot analysis.

Abstract

The scatter plot is a commonly used assessment tool for identifying temporal patterns in the occurrence of behavior problems. However, the extent to which such patterns are frequently observed is unknown because little research has evaluated the general utility of the scatter plot. We conducted a large-scale analysis of within- and across-day occurrences of problem behavior by conducting continuous observations of 20 individuals living in four residential facilities. Data were recorded during 30-min intervals throughout participants' waking hours for 30 days by direct care staff and were converted into scatter plot formats. Five sets of data were excluded from further analysis due to poor interobserver agreement (below 80%). Visual analysis of the remaining 15 scatter plots indicated that none showed any reliable temporal pattern of responding. However, when the data were transformed into aggregate "control charts" based on statistical process control procedures, 12 of the 15 sets of data revealed one or more 30-min intervals during which problem behavior was more likely to occur. Results are discussed in terms of the practicality of applying statistical analyses to scatter plot data and of collecting data for the length of time needed to show statistical significance. It was concluded that detailed functional or descriptive analyses, which would reveal cause-effect or correlational relationships between behavior and specific environmental events, may be both more precise and more efficient forms of assessment.