A Permutation Test for Unbalanced Paired Comparisons of Global Field Power

Area of Science:

• Neuroscience
• Cognitive Science
• Biostatistics

Background:

• Global Field Power (GFP) is a key metric summarizing multi-channel electroencephalography (EEG) data.
• Standard GFP analysis is susceptible to noise, invalidating comparisons when EEG datasets have unequal noise levels.
• Unequal numbers of trials across conditions introduce significant bias in traditional GFP comparisons.

Purpose of the Study:

• To elucidate the relationship between the number of trials and the expected value of GFP.
• To introduce a novel statistical testing procedure for valid within-subject comparisons of GFP with unequal trial counts.
• To address the limitations of conventional methods in handling unbalanced multi-subject EEG data.

Main Methods:

• Demonstrated the mathematical relationship between trial count and expected GFP.
• Developed and validated a new statistical test for repeated-measures GFP analysis with unequal trials per condition.
• Utilized simulations to compare the proposed test against conventional approaches under unequal trial conditions.
• Applied the proposed test and alternatives to empirical data from a visual target detection experiment.

Main Results:

• Simulations confirmed the proposed test's validity and highlighted the invalidity of conventional methods with unbalanced data.
• The new test successfully identified significant GFP differences in the P3 range in experimental data.
• Alternative tests detected differences but also yielded spurious findings outside the expected temporal range.

Conclusions:

• The proposed statistical test effectively corrects for trial number bias in GFP analysis.
• This method enables sensitive and valid within-subject comparisons of GFP in multi-subject, unbalanced EEG datasets.
• The findings support the use of the new procedure for more reliable EEG data interpretation.