NMAK14015U Multivariate analysis (MultivAn)

Multivariate statistics is about the statistical analysis of data with more than one response variable. The collection of response variables may either have rather low dimension, from 2 to 7, say, or be genuine high dimensional. Analyzing multivariate data can be computationally demanding, and for many of the in practice applied methods there is a strong emphasis on the algorithmic aspects. Such methods includes PCA (principal component analysis), factor analysis, PLS (partial least squares) and PARAFAC (parallel factor analysis). We will describe these methods, and the data situations for which they apply. At the same time we will look for statistical models leading to the listed estimation methods, and see to which degree the multivariate methods can be extended to and embedded in classical statistical models like repeated measurements. As a restriction we exclusive will deal with responses that somehow can be described by a multivariate Gaussian distribution.