Abstract

The utility of a practical or other measure can be assessed in a validity study, in which values of the measure are compared with those of a criterion measure taken concurrently in a sample of subjects. A scatterplot of criterion vs practical values provides a qualitative assessment of non-linearity, random error and systematic error in the relationship between the two measures, while the statistics of linear regression (equation of the line or curve, standard error of the estimate, correlation coefficient) provide not only a quantitative assessment but are also useful for interpreting and adjusting values and effects involving the practical measure. Another method for comparing two measures, suggested by Bland and Altman, is based on a plot and analysis of the difference between the measures. Although in widespread use, the Bland-Altman method is inappropriate for validity studies: the plot shows systematic error incorrectly and the assessment of interchangeability does not properly reflect the utility of the practical measure. If there is no criterion measure in a measure-comparison study, use of regression or Bland-Altman approaches is pointless without a strategy to rank the measures. Comparison of correlation coefficients between a sufficient number of measures is one such strategy. Verf.-Referat