Schröder, Felix (2021) The Effects of Autocorrelation on Overlap Corrected rERPs. Research Project 2 (major thesis), Behavioural and Cognitive Neurosciences.
|
Text
FSchroder_Mthesis_Final_s3086402.pdf Download (736kB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (99kB) |
Abstract
Regression-based deconvolution allows for the separation of overlapping event related potentials. This carries great potential for naturalistic study designs but has not been validated for within subject analyses so far. As opposed to across subject analyses, tests on within subject estimates may not be valid since t-tests on individual β estimates rely on the assumption of independence to allow inference. EEG data violates this assumption. In this report we assessed the effects of this non-independence on type-1 error rates of β parameters estimated from EEG via continuous-time regression. To this end, we simulated null-signals that had different degrees of non-independence. Subsequently, a number of different experimental conditions were simulated, modelled, and their parameters tested. We found non-independence of the residuals increased type 1 errors dramatically once systematic overlap was added to the model. We further discuss possible solutions that could be tested. Since EEG data violate the assumption of independence to a much greater extent than traditionally dealt with in the time-series literature, none of these solutions should be trusted without thorough empirical testing beforehand. Thus, we discourage the application of overlap-correction for within-subject analyses for now.
| Item Type: | Thesis (Research Project 2 (major thesis)) |
|---|---|
| Supervisor name: | Mathot, S. |
| Degree programme: | Behavioural and Cognitive Neurosciences |
| Thesis type: | Research Project 2 (major thesis) |
| Language: | English |
| Date Deposited: | 06 Oct 2021 14:57 |
| Last Modified: | 06 Oct 2021 14:57 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/26173 |
Actions (login required)
 |
View Item |