The Devil Rides Again: Correlation as an Index of Fit

Correlations between theoretical predictions and data can be higher for incorrect than for correct models, as illustrated by analyses of two sets of hypothetical data. This fact raises questions about the conclusions of recent studies that use correlation as an index of fit. Functional measurement provides a sounder basis for model evaluation by placing scaling in the context of model fitting and by testing deviations from prediction rather than concent rating an overall goodness of fit.