Methodological improvement for the statistical analysis of neurophysiological data collected from informations systems

点击翻译按钮获取中文摘要

This research program stems from the various research projects of the Tech3Lab, a research laboratory who specializes in the use of multi-method approaches for measuring user's experience. At the Tech3lab, we jointly use an eye tracker, electroencephalography and devices collecting many physiological measures (respiration, heart rate, galvanic skin response and facial emotions) while one or many subjects are interacting with a computer to perform a specific task (visiting a website, playing video games, doing work-related tasks on a software, etc.). This new possibility to jointly analyze these types of complex data creates the need to develop new statistical models, and more efficient statistical techniques, adapted to better answer research problems specific to our field. For instance, in event-related potential (ERP) or eye fixation-related potential (EFRP) experiments, the noisy signals coming from electroencephalography data are difficult to analyze. In traditional neuroscience research, this problem is usually solved by collecting data on a very large number of ERP/EFRP trials in order to average out the noise. However, this is often not possible for research problems were we want to preserve the realism of an experiment. The need to develop efficient statistical methods to explicitly test EEG data with a limited number of ERP/EFRP trials is thus clear. Another situation where more efficient statistical techniques will be useful is in the joint modelling of EEG data collected simultaneously on different subjects. A more precise statistical analysis of the coherence between brains will be useful to study concepts like the apparition of flow in multi-player video games or teamwork performance.