Specification dependence of EEG theta/alpha ratio associations with cognitive impairment: a multiverse analysis

The study investigates the impact of various preprocessing, spectral, spatial, and statistical decisions on the associations between EEG theta/alpha ratio and cognitive impairment. By conducting a multiverse analysis on two resting-state EEG datasets, the researchers aimed to evaluate the robustness of these relationships across a wide range of analytical pipelines.

The first dataset comprised 49 healthy controls and 100 participants with Parkinson's disease, ranging from normal cognition to dementia. The second dataset consisted of 29 healthy controls, 36 patients with Alzheimer's disease, and 23 patients with frontotemporal dementia. Eight decisions were varied across the different analytical universes, including filtering, artifact correction, independent component analysis, referencing, spatial summary, alpha-band definition, spectral method, and covariate adjustment. This resulted in a total of 1,280 possible analytical specifications, or universes.

The findings of the multiverse analysis revealed significant variability in the results depending on the specific combination of decisions employed. In Dataset 1, the exploratory analysis family, which combined ICA with either absolute fast Fourier transform (FFT) or Welch power spectral density estimation, yielded statistically significant p-values indicating a relationship between the theta/alpha ratio and cognitive impairment.

However, the authors caution that such associations may be sensitive to the specific preprocessing and analytical choices made, and further research is needed to establish the true magnitude and consistency of this relationship.