A Model-Based Framework for Assessing the Physiologic Structure of Electrodermal Activity

Objective: We present a statistical model for extracting physiologic characteristics from electrodermal activity (EDA) data in observational settings. Methods: based our on the integrate-and-fire physiology of sweat gland bursts, which predicts inverse Gaussian (IG) inter-pulse interval structure. At core model-based paradigm is subject-specific amplitude threshold selection process EDA pulses properties four right-skewed models including IG. By performing sensitivity analysis across thresholds and fitting all models, we selected IG-like structure verified pulse with goodness-of-fit analysis, maximizing capture at time scale responses. Results: tested simulated series two different experimental cohorts recorded during conditions, using equipment. In both data, method robustly recovered that captured predicted by physiology, despite large differences noise level. contrast, established tools, attempted to estimate neural slower responses, did not provide physiological validation were susceptible noise. Conclusion: computationally efficient, statistically rigorous, physiology-informed robust individuals yet adaptable varying Significance: The robustness its basis empirical support use as clinical marker sympathetic conditions such pain, anxiety, depression, sleep states.