University of Twente Student Theses

Login

Statistical Analysis of EEG (Trial) Data through Covariance Structure Modelling

Bazen, J.L. (2023) Statistical Analysis of EEG (Trial) Data through Covariance Structure Modelling.

[img] PDF
501kB
Abstract:The analysis of EEG (trial) data is an intensive procedure. Multiple trials will have to be measured to account for random error. The trial data are often averaged, and event related potentials (ERPs) need to be identified. Then, statistical models like the linear mixed effects model (LME) are used for statistical inferences. A good fit of the LME requires support of the factor structure by the data, which can be a challenging task. Furthermore, the random factor variances are restricted to be positive, which limits its suitability for complex structured data. Bayesian Covariance Structure Model (BCSM) bypasses this limitation by modelling directly the correlational structure in the data. A simulation study was conducted for the evaluation of the BCSM compared to the LME model for different sample sizes and covariance values. Then, the BCSM was applied to electroencephalogram (EEG) trial data. The results of the simulation study indicated good performance of the BCSM, for the higher covariance values the models showed comparable performance. For the lower and negative covariances the BCSM performed better. The BCSM showed good recovery of the 95% (posterior) credible intervals and acceptable (posterior) standard deviations. Lastly, the BCSM provided plausible results for the fit on the EEG data. However, as there is no benchmark for quality, any conclusions should be drawn with care. Summarized, this study indicates a promising future for the BCSM, if a way to validate its results can be established.
Item Type:Essay (Bachelor)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:70 social sciences in general, 77 psychology
Programme:Psychology BSc (56604)
Link to this item:https://purl.utwente.nl/essays/95279
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page