Misallocation of variance in event-related potentials: simulation studies on the effects of test power, topography, and baseline-to-peak versus principal component quantifications

Abstract

Since Wood and McCarthy's simulation study (Electroenceph Clin Neurophysiol 1984;59:249-260), the use of principal component analysis (PCA) as a tool for the identification and quantification of event-related potentials (ERP) has been considered a challenge. Three relevant aspects have not been fully acknowledged in previous studies, however, and were therefore investigated in the present simulation study. Firstly, the impact of test power on the amount of variance misallocation was studied. Secondly, the impact of ERP component topography on variance misallocation was investigated. Thirdly, a systematic evaluation of variance misallocation in baseline-to-peak derived ERP measures was performed. Results based on an overall set of 2700 simulations indicate that: (a) variance misallocation is reduced to an almost acceptable level when an appropriate test power is simulated; (b) the overall amount of variance misallocation remains at an almost acceptable level when systematic topographic effects are simulated in combination with an appropriate test power; and (c) variance misallocation is in fact also a problem in baseline-to-peak measures. These findings confirm that, when used appropriately, PCA is a helpful and efficient tool for the identification and quantification of ERPs.