100 years of multivariate analysis

The date of birth of a field is seldom well-defined. More usually there are several key events from which a field then grows. In the case of multivariate analysis, a distinguishing feature is the relation between variables, and how are dependencies measured. Here the key ingredient was the correlation coefficient introduced by Karl Pearson. At that time there were no inferential methods, and here the year 1908 is important (exactly 100 years ago) with the introduction of Studentpsilas t-statistic. This was a univariate procedure, but it was needed before the next step could be taken. Thus from this point on multivariate procedures began to be developed. John Wishart obtained the distribution of the sample covariance matrix, and this was a central missing link for almost all future developments. Although new multivariate methods have been developed over the years, more recently there has been a resurgence of multivariate procedures in finance and in data mining. In this talk I will walk through some of the main ideas that have been the focus of research in multivariate analysis.