University of Twente Student Theses


Testing the successive time window approach with simulated EEG data

Coroiu, A. (2023) Testing the successive time window approach with simulated EEG data.

[img] PDF
Abstract:The multiple testing problem arises when trying to interpret electroencephalogram (EEG) recordings with no a priori assumptions about the timing or location of the expected effect. Noise and data dependencies increase the probability of observing false positives in highly dimensional EEG data. Therefore, the family wise error rate (FWER) needs to be controlled through appropriate statistical methods. In this study, the performance of the successive time window (STW) approach was tested. The specificity, sensitivity, and precision of this approach were assessed for identifying effects on visual event-related potentials (ERPs). The results of this approach were compared to other popular FWER correction methods. The MNE python library was used to create fully synthetic EEG data for a Monte Carlo simulation. Different data parameters were used to define EEG datasets for a between-condition experiment with visual ERPs from 20 simulated subjects. The data were analysed using the STW approach, classic Bonferroni, and a cluster permutation (CP) method. Local and global type I, and II error rates (ERs), and the false discovery rate (FDR) were used to quantify the performance of these methods. The results of this study show that the STW approach leads to a lower local type II ER compared to Bonferroni, but a higher global type I ER (FWER) when compared against both other methods. The STW approach and Bonferroni provide a similar FDR, with better resolution than the CP method. Therefore, the STW approach does not control the FWER as well as the CP and Bonferroni methods, but it can provide more sensitivity and precision. The performance of the STW approach can be highly improved when some a priori assumptions can be made about the location of an expected ERP effect. The critical p-value calculation proposed for the STW approach can be improved by adjusting it according to the noise level in rest state baseline EEG data. Future research can build on the methodology of this study to further validate statistical methods aimed at solving the multiple testing problem in ERP studies.
Item Type:Essay (Master)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:77 psychology
Programme:Psychology MSc (66604)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page