Visual Inspection and Statistical Analysis in Single-Case Designs

Abstract

Past research has questioned the reliability of visual inspection in single-case research. It has also reported or called for the use of statistical analysis in single-case research. The present study randomly selected 44 graphs from recent issues (1975–1985) of the Journal of Applied Behavior Analysis. The study found moderate agreement (56–74%) of visual significance among a panel of five judges. The judges also detected a rather low proportion (48%) of significant differences between the adjacent phases of published graphs. A randomization test was applied to a subset (n = 15) of the graphs and found statistical significance between only four of the adjacent phases. Considerable agreement (80%) was found between the visual inspection and statistical procedures, although most of it involved agreement of the nonsignificance of treatment effects. It was recommended that future single-case research jointly utilize visual and statistical procedures when determining significance. It was also suggested that future research in the field examine graphs with multiple phases, baselines, and participants.