ABSTRACT
This chapter provides an overview of the relationships between the most common and the most powerful approaches to analyzing neuroimaging data. It also provides new practitioners with the intuition and technical knowledge required to apply multivariate statistical techniques, many of which have been developed in or adapted from the field of machine learning. The general linear model (GLM) is a tireless workhorse of cognitive neuroimaging research. One designs an experiment around a GLM analysis with the hope that it will reveal a brain region that is more activated by one sensory, behavioral, or cognitive state—known in the GLM world as “conditions”— than some other. The chapter suggests that cognitive neuroscience in its current stage of growth has settled on a strategy of linearization. However, as long as cognitive neuroscience continues to be guided by the assumption of pan-linearity then it is straightforward to predict how machine-learning techniques will influence and interact with cognitive neuroscience.
