Maximizing signal-to-noise ratio and statistical power in ERP measurement: Single sites versus multi-site average clusters.

Abstract

One important decision in every event-related potential (ERP) experiment is which electrode site(s) to use in quantifying the ERP component of interest. A common approach is to measure the ERP from a single electrode site, typically the site where the ERP component is largest. Alternatively, two or more electrode sites in a given spatial region are averaged together, and the ERP is measured from the resulting multi-site cluster. The goal of the present study was to systematically compare these two measurement approaches across a range of outcome measures and ERP components to determine whether measuring from a single electrode site or an average of multiple sites yields consistently better results. We examined seven common ERP components from the open-source ERP CORE dataset that span a range of neurocognitive processes: N170, mismatch negativity (MMN), N2pc, N400, P3, lateralized readiness potential (LRP), and error-related negativity (ERN). For each component, we compared ERP amplitude, noise level, signal-to-noise ratio, and effect size at two single electrode sites and four multi-site clusters. We also used a Monte Carlo approach to simulate within-participant and between-groups experiments with known effect magnitudes to compare statistical power at single sites and multi-site clusters. Overall, measuring from a multi-site cluster produced results that were as good as or better than measuring from a single electrode site across analyses and components, indicating that the cluster-based measurement approach may be beneficial in quantifying ERPs from a range of neurocognitive domains.