Electroencephalographic fingerprints of statistical model selection by the brain

• Ano de defesa: 2022
• Autor(a) principal: Najman, Fernando Araujo
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Cérebro estatístico; Electroencephalography; Eletroencefalografia; Seleção estatística de modelos; Statistical modelselection; Statistician brain
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/95/95131/tde-02032022-151157/

It has been classically conjectured that the brain compresses data by assigning probabilistic models to sequences of stimuli. An important issue associated with this conjecture is to identify classes of models used by the brain to perform its compression task. We address this issue by introducing two new statistical methods for functional data. The first method is a statistical model selection procedure for EEG data recorded using the context tree paradigm. This methodology retrieves a dendrogram model from EEG data. We present new results with both simulated and experimental EEG data obtained using this methodology which show that our method can discriminate between different conjectures about which information of the stimuli sequence is encoded in the electrical activity of the brain even if the information is not well described by a context tree model. The second method is a new graphical procedure for functional data. This procedure gives us a visual representation of the time domain in which two sets of functional data differ. We discuss how this graphical procedure can be used as an exploratory tool to better understand the inter-individual variance observed in experimental EEG data and how different participants encode the stimuli sequence information.