Published online by Cambridge University Press: 29 January 2018
In studies which attempt to assess the efficacy of some new treatment it is customary to make a formal assessment of the relevant behaviour or symptomatology using standardized inventories, checklists, symptom rating scales or ad hoc ratings of variables considered to be important by the clinician-researcher. Ratings made at the beginning and end of treatment are compared for groups of patients using improvement scores, arbitrary cut-off points and other devices aimed at circumventing the statistical problems arising from the non-normal distributions of most rating scales. Present practice favours the use of some nonparametric statistic in these comparisons. The aim of this paper is to present a method which facilitates the analysis of ratings made on more than two occasions, allowing a trend analysis to be carried out without making assumptions about the distribution of scores. The method also allows the clinician-researcher to make a statistically-based decision regarding the efficacy of the treatment in question for individual patients.
eLetters
No eLetters have been published for this article.