Beyond the Mean: Can We Improve the Predictive Power of Psychometric Scales?

Abstract

Two participants completing a psychometric scale may leave wildly different responses yet attain the same mean score. Moreover, the mean score often does not represent the bulk of participants’ responses, which may be skewed, kurtotic, or bimodal. Even so, researchers in psychological science often aggregate item scores using an unweighted mean or a sum score, thereby neglecting a substantial amount of information. In the present contribution, we explore whether other summary statistics of a scale (e.g., the standard deviation, the median, or the kurtosis) can capture and leverage some of this neglected information to improve prediction of a broad range of outcome measures: life satisfaction, mental health, self-esteem, counterproductive work behavior, and social value orientation. Overall, across 32 psychometric scales and three data sets (total N = 8,376), we show that the mean is the strongest predictor of all five outcomes considered, with little to no additional variance explained by other summary statistics. These results provide justification for the current practice of relying on the mean score but hopefully inspire future research to explore the predictive power of other summary statistics for relevant outcomes. For this purpose, we provide a tutorial and example code for R.