On the Application of Multivariate Statistical and Data Mining Analyses to Data in Neuroscience.

Research in neuroscience, whether at the level of genes, proteins, neurons or behavior, almost always involves the interaction of multiple variables, and yet many areas of neuroscience employ univariate statistical analyses almost exclusively. Since multiple variables often work together to produce a neuronal or behavioral effect, the use of univariate statistical procedures, analyzing one variable at a time, limits the ability of studies to reveal how interactions between different variables may determine a particular outcome. Multivariate statistical and data mining methods afford the opportunity to analyze many variables together, in order to understand how they function as a system, and how this system may change as a result of a disease or a drug. The aim of this review is to provide a succinct guide to methods such as linear discriminant analysis, support vector machines, principal component and factor analysis, cluster analysis, multiple linear regression, and random forest regression and classification, which have been used in circumscribed areas of neuroscience research, but which could be used more widely.