When Is A Picture Worth A Thousand $Rh Values? A Comparison Of Visual And Quantitative Methods To Analyze Single Subject Data

The agreement between visual analysis and the results of the split-middle method of trend estimation was examined using a set of 24 stimulus graphs. Thirty raters indicated whether a significant change occurred across the phases of the stimulus graphs. The average visual analysis score for each graph was then compared to the results of the split-middle method of trend estimation. Using the trend line and tables based on the cumulative binomial probability distribution, a statistical statement of change across design phases was generated. The level of agreement between visual and statistical inferences was .46. The sensitivity, specificity, and predictive value of visual analysis in relation to the statistical conclusions ranged from .38 to .73. Findings indicate the need for continued investigation of the various properties of visual and statistical analysis as applied to single subject data.