EVALUATION OF SINGLE-SUBJECT DATA USING STATISTICAL HYPOTHESIS TESTS VERSUS VISUAL INSPECTION OF CHARTS WITH AND WITHOUT CELERATION LINES

The accuracy of visual analysis of single subject charts with and without celeration lines was compared. A Monte Carlo simulation was used to create five types of AB design charts: (I) a flat (zero slope) baseline with no systematic change in intervention phase, (II) a flat baseline followed by a one standard deviation increase in level in intervention phase, (III) an ascending baseline with intervention phase responses following the baseline phase trend (no systematic change), (IV) an ascending set of baseline and intervention responses with a one standard deviation decrease in level in the intervention phase, and (V) an ascending baseline followed by descending intervention phase responses. These charts were evaluated by 271 members of the Association for Advancement of Behavior Therapy. Charts with celeration lines were evaluated by 140 subjects and charts without celeration lines were evaluated by 131 subjects. Greater decision accuracy was associated with presence of celeration lines for situations where there was a trend in the data and where a change in level or slope had occurred between phases (conditions IV and V). Otherwise, there were no statistically significant differences in decision accuracy between evaluations of charts with celeration lines versus without celeration lines. Decision accuracy of statistical hypothesis tests was evaluated Utilizing either a t-test for mean differences or piecewise regression (depending upon the results of an evaluation for trend within phases) was more accurate than visual inspection with celeration lines in situations where no change had occurred between phases (conditions I and III). Other differences were not statistically significant.