Using Robust Statistics and Distribution Parameters to Establish Valid Individual Differences in Computer-Based Cognitive Testing

Abstract

Data gathered in large-scale computer-based cognitive laboratories occasionally appear less orderly than data gathered on similar tasks in laboratories where close interaction between subject and experimenter is possible. The present research addressed the question of whether the use of robust statistics could make such data appear more orderly and lawful. Of related interest was the question of whether parameters which describe whole distributions of data, rather than only means and standard deviations, could be used to establish reliable and valid individual differences. For the large sample sizes used here the robust estimators did not outperform standard statistics. The distribution parameters were not sufficiently reliable to be useful in characterizing psychological processes.