Multimodal diffusion model for increments of electroencephalogram data

Abstract We propose a new strictly stationary strong mixing diffusion model with marginal multimodal (three-peak) distribution and exponentially decaying autocorrelation function for modeling of increments electroencephalogram data collected from Ugandan children during coma cerebral malaria. treat the as discrete-time observations construct process where is viewed mixture three non-central generalized Gaussian distributions we state some important properties related to moments this mixture. estimate parameters using expectation-maximization algorithm, added shape parameter estimated higher order statistics approach based on an analytical relationship between kurtosis. The derived estimates are then used prediction subsequent neurodevelopment cognition malaria survivors elastic net regression. compare different predictive models determine whether additional information obtained multimodality can be improve prediction.