Parametric vs. non-parametric statistics of low resolution electromagnetic tomography (LORETA)

Area of Science:

• Neuroscience
• Biostatistics

Background:

• Accurate statistical analysis of 3D current sources is crucial for interpreting electroencephalography (EEG) data.
• Low Resolution Electromagnetic Tomography (LORETA) is a common EEG inverse solution for estimating these sources.
• Comparing the statistical sensitivity of parametric and non-parametric methods is essential for optimizing LORETA analysis.

Purpose of the Study:

• To compare the statistical sensitivity of parametric and non-parametric methods for analyzing 3D current sources estimated by LORETA.
• To determine the optimal statistical approach for minimizing false positives in LORETA-based EEG analysis.

Main Methods:

• EEG data from 43 normal adults were processed using the Key Institute's LORETA program.
• A "leave-one-out" cross-validation method was employed to assess classification accuracy.
• Log10 transforms were applied to approximate Gaussian distributions for statistical testing.

Main Results:

• Parametric Z-score tests with a log10 transform achieved high accuracy (95-99%) in approximating Gaussian distributions.
• At P < .01, parametric Z-score tests yielded a low false positive rate of 0.26%.
• Non-parametric t-max statistics showed higher average misclassification error rates (7.64% at P < .05, 6.67% at P < .01) compared to parametric tests.

Conclusions:

• Log10 transformation and parametric statistics provide adequate approximation to Gaussian distribution for LORETA analysis.
• Parametric normative comparisons demonstrate lower false positive rates than non-parametric tests.
• The findings support the use of parametric statistics with log10 transforms for robust EEG source analysis using LORETA.