Gosselin Frédéric, Larouche Jean-Maxime, Daigneault Valérie, Caplette Laurent
Département de Psychologie, Université de Montréal, Montréal, Canada.
Department of Psychology, Yale University, New Haven, CT, USA.
Behav Res Methods. 2024 Mar;56(3):2452-2468. doi: 10.3758/s13428-023-02158-6. Epub 2023 Jul 10.
This paper introduces a novel procedure that can increase the signal-to-noise ratio in psychological experiments that use accuracy as a selection variable for another dependent variable. This procedure relies on the fact that some correct responses result from guesses and reclassifies them as incorrect responses using a trial-by-trial reclassification evidence such as response time. It selects the optimal reclassification evidence criterion beyond which correct responses should be reclassified as incorrect responses. We show that the more difficult the task and the fewer the response alternatives, the more to be gained from this reclassification procedure. We illustrate the procedure on behavioral and ERP data from two different datasets (Caplette et al. NeuroImage 218, 116994, 2020; Faghel-Soubeyrand et al. Journal of Experimental Psychology: General 148, 1834-1841, 2019) using response time as reclassification evidence. In both cases, the reclassification procedure increased signal-to-noise ratio by more than 13%. Matlab and Python implementations of the reclassification procedure are openly available ( https://github.com/GroupeLaboGosselin/Reclassification ).
本文介绍了一种新颖的程序，该程序可在以准确性作为另一个因变量的选择变量的心理实验中提高信噪比。此程序基于这样一个事实：一些正确反应是猜测的结果，并使用诸如反应时间之类的逐次试验重新分类证据将它们重新分类为错误反应。它选择最佳的重新分类证据标准，超过该标准正确反应应被重新分类为错误反应。我们表明，任务越困难且反应选项越少，从这种重新分类程序中获得的收益就越大。我们使用反应时间作为重新分类证据，在来自两个不同数据集（Caplette等人，《神经影像学》218卷，第116994页，2020年；Faghel-Soubeyrand等人，《实验心理学杂志：总论》148卷，第1834 - 1841页，2019年）的行为和ERP数据上说明了该程序。在这两种情况下，重新分类程序使信噪比提高了超过13%。重新分类程序的Matlab和Python实现可公开获取（https://github.com/GroupeLaboGosselin/Reclassification ）。
