Research › QEEG, Methods & Mechanisms
Parametric vs. Non-Parametric Statistics of Low Resolution Electromagnetic Tomography (LORETA)
Share
Study at a glance
- Area
- QEEG, Methods & Mechanisms
- Study type
- Observational study
- Authors
- Thatcher RW, North D, Biver C
- Journal
- Clinical EEG and Neuroscience
- Year
- 2005
- Participants
- 43 healthy adults
- Population
- Adults
- Protocol
- eyes-closed resting EEG, LORETA leave-one-out cross-validation
Key findings
- Log10 transforms approximated a Gaussian distribution with 95% to 99% accuracy
- Parametric Z-scores misclassified about 4.25% of normal subjects at P < .05 and 0.26% at P < .01
- Non-parametric t-max misclassified an average of 7.64% at P < .05 and 6.67% at P < .01
- Parametric comparisons had lower false positive rates than non-parametric tests
This methods paper compared parametric and non-parametric statistics for deciding whether a person's low-resolution electromagnetic tomography (LORETA) current source values are abnormal. Eyes-closed EEG from 43 healthy adults was processed with the Key Institute LORETA software, and each person was left out in turn and classified against the others to estimate false positive rates. After a log10 transform, parametric Z-scores gave fewer false positives than the Key Institute's non-parametric t-max test. The authors conclude parametric normative comparisons work well with LORETA.