Donchin E
Department of Psychology, University of Illinois, Urbana 61920.
Biol Psychol. 1989 Apr;28(2):181-6. doi: 10.1016/0301-0511(89)90100-2.
Collet's (1989) doubts regarding the efficacy of principal component analysis (PCA) as a tool in the study of event-related brain potentials (ERPs) are unpersuasive. The substantive point reported by Collet is that data points are autocorrelated over the time period during which a single, PCA-defined, ERP component predominates. Such autocorrelation has long been recognized by investigators in the field, and its confirmation by autoregressive modeling does not provide useful information about issues central to ERP data analysis. Furthermore, because he fails to take a full count of the number of parameters used in his autoregressive model his argument from parsimony is flawed. In any event, Collet's argument misperceives the heuristic mode in which PCA is used in actual studies.
科莱（1989年）对主成分分析（PCA）作为研究事件相关脑电位（ERP）的工具的有效性表示怀疑，这种怀疑缺乏说服力。科莱报告的关键问题是，在单个由PCA定义的ERP成分占主导的时间段内，数据点存在自相关。该领域的研究人员早就认识到这种自相关，并且通过自回归建模对其进行确认并不能为ERP数据分析的核心问题提供有用信息。此外，由于他没有完全计算出自回归模型中使用的参数数量，他基于简约性的论点存在缺陷。无论如何，科莱的论点误解了PCA在实际研究中的启发式模式。
