Evaluation of individual subjects in the analog classroom setting: I. Examples of graphical and statistical procedures for within-subject ranking of responses to different delivery patterns of methylphenidate.

In this article, we describe graphical and statistical methods developed to evaluate the response patterns of individual children with attention deficit hyperactivity disorder (ADHD) to different conditions of treatment with stimulant medication. We used data from an investigation of drug delivery patterns to demonstrate these methods. Thirty-one children with ADHD participated in a double-blind crossover study of four conditions (three patterns of delivery of methylphenidate and a placebo control). In each condition, the children were evaluated across an 11-hour (7:00 a.m. to 6:00 p.m.) laboratory school day, and ratings of classroom behavior were obtained at regular intervals across the day. Graphical procedures were developed to display, for each individual, time courses of multiple measures of behavior taken across each double-blind test day. Expert clinicians judged these graphs and used this information to rank-order the test days from best to worst. A within-subject variant of Kendall's W was used to evaluate, for each subject, whether the rankings of these multidimensional graphs were reliable (concordant) across judges. A generalized kappa statistic was used to evaluate, for each condition, the reliability of the judges' rankings across subjects. Friedman's analysis of variance of ranks was used to evaluate, for the study, whether the conditions differed in terms of the average (consensus) rank assigned by the judges.