Key points are not available for this paper at this time.
The purpose of this study is to examine the implications of various less-examined psychometric issues in the interpretation of neuropsychological data. Using a dataset of 4371 independent functioning and community-dwelling individuals who underwent neuropsychological evaluations, it was demonstrated that many common measures are not normally distributed. Non-normalized data can lead to erroneously pathological conclusions, particularly on the lower end of negatively skewed distributions. Another issue involves scatter. In line with previous studies, the current study found that approximately 67% of the 4371 participants showed discrepancies of three or more standard deviations between their highest and lowest test scores on 21 measures. However, in contradiction to the existing literature, in the current study mean scatter levels were relatively stable across increasing levels of intelligence. It is argued that this is due to regression to the mean. As an individual moves away from the population average in either direction, scores on other measures will regress from that person's IQ score toward the population mean. The lower a test correlates with IQ, the greater will be the regression toward the mean. Therefore the test battery in question must be considered during the interpretation process, in addition to the individual's premorbid IQ.
Donnell et al. (Tue,) studied this question.