A regression-based framework for estimating and controlling for trial-to-trial variations in behavioral measurements

Repeated measurements, such as trials on a task, are a ubiquitous feature of behavioral studies. These repeated measures may be influenced by many time-dependent phenomena such as learning, adaptation, fatigue, or mind-wandering. Such phenomena may be of theoretical importance or may be nuisance factors, but in both cases researchers would benefit from estimating (and possibly controlling for) time-dependent changes. Because methods for estimating and/or controlling for time-dependent variations have been limited, here we demonstrate a method of augmenting standard regression-based analyses to provide sensitivity to variations over time. Through two sets of simulations, we show the applicability and implementation of this method for several types of data and research domains. We then show, in empirical data from a perceptual decision-making study, that our approach can recapitulate the results of theory-driven analyses of changes over time.