Abstract

Background: Neuropsychological tests (NTs) with different number of items (test length), number of response-categories/levels (width), dimensions covered, different values of reliability, validity, responsiveness, discriminating value, etc. are not comparable. Aim: To address methodological issues of neuropsychological testing and suggest remedial measures by transforming discrete item scores to normally distributed continuous scores for meaningful evaluation of measurement properties and better utilization of such tests. Methods: Using data driven weights to different response-categories of different items, ordinal item scores are converted to equidistant score (E-scores) in ratio scale with fixed zero point. Proposed scores (P-scores) are obtained via standardization of E-scores followed by further transformation to follow normal distribution. Sub-class/dimension scores and test scores are obtained as sum of item-wise P-scores. Results: Normally distributed P-scores facilitate arithmetic aggregation with cardinal measures like number of errors, time taken, etc. and offer platform for parametric analysis including statistical testing. The proposed P-scores also help to find theoretically defined reliability, factorial validity avoiding criterion scale, discriminating value, assessment of progress/deterioration, efficiency of classification, equivalent scores of neuropsychological tests. Conclusion: Proposed scores following normal distribution and satisfying desired properties of measurement is recommended. Practicing psychiatrists and researchers can derive benefits of the proposed score for more meaningful comparisons, classifications, equating boundary scores for classification, and assessment of progress or deterioration.

Keywords

neuropsychological test battery, normal distribution, ratio scale, equivalent scores, theoretical reliability, factorial validity