Objectives: After adjustment of scores for demographic variables, especially when test scales have an upper limit that causes a ceiling effect, the original score variability is deeply altered and the scale properties degenerate. We present a method for fixing normality thresholds on scores previously adjusted for demographic variables that overcomes these problems.
Methods: We suggest to fix norms using non-parametric tolerance limits. These limits, valid even for asymmetrical distributions and adjusted scores, can also be used as a basis for a non-parametric standardization of adjusted scores. Non-parametric tolerance limits permit to fix, with controlled risk, the threshold under which there is at most 5% of the normal population (outer tolerance limit); this approach also permits to fix, still with a stringent risk control, the inner tolerance limit, i.e. the threshold under which there is at least 5% of the normal population. Separate limits are needed to control not only against falsely declaring that an individual is 'not normal' (outer tolerance limit), but also against falsely declaring that an individual is 'normal' (inner tolerance limit).
Results: We provide examples of the calculations necessary and some tables useful for clinical practice, both for the normality judgment and for a standardization method called Equivalent Scores.
Conclusions: This approach has been followed for about 80 neuropsychological tests studied in the Italian population. Besides the intrinsic value of this method, its extension to a wider audience of neuropsychologists could provide a way to gain from the experimental and clinical practice carried out in different countries.
Keywords: Tolerance limits; demographic adjustment; outer and inner limits; standardization.