Estimating statistical power for event-related potential studies using the late positive potential

Area of Science:

• Cognitive Neuroscience
• Affective Neuroscience
• Psychophysiology

Background:

• The late positive potential (LPP) is crucial for studying emotional processing in electroencephalography (EEG) research.
• Current research lacks a standardized method for determining adequate statistical power in LPP studies.
• Understanding factors influencing LPP statistical power is essential for robust experimental design.

Purpose of the Study:

• To investigate the impact of trial count, subject numbers, and effect size on statistical power for LPP analysis.
• To provide empirical data for optimizing ERP study designs focusing on the LPP component.

Main Methods:

• Utilized Monte Carlo simulations to generate 1,489 unique ERP experiment scenarios.
• Varied the number of trials, subjects, and synthetic effect sizes (magnitude) within simulations.
• Repeated each simulation 1,000 times to calculate the probability of achieving statistical significance.

Main Results:

• Statistical power consistently increased with more trials, subjects, and larger effect sizes.
• Within-subject designs demonstrated higher statistical power compared to between-subjects designs, particularly with fewer trials/subjects and smaller effects.
• Adding subjects significantly boosted statistical power at lower effect sizes (<1 µV), with diminishing returns at larger effect sizes (>1.5 µV).

Conclusions:

• Trial and subject numbers, alongside effect size, are critical determinants of LPP study power.
• Within-subject designs are more power-efficient for LPP research than between-subjects designs.
• Simulation results based on synthetic data were corroborated by simulations using real EEG data from emotional image viewing.