Unit IV - Models of statistical dependency

Italian version

Lecturer: Marcello Chiodi

We present some statistical basic model to study dependence between variables in multivariate data sets. PCA (Principal Component Analysis) is recalled as a basic technique for the study of multivariate interdependence and collinearity. General Linear models are recalled together with some issues related to choice of models and classical variance-bias trade off (cross validation and AIC are presented in real data application. Some extension are presented for non independent errors (time models and space-time models), and GLM (Generalized Linear Models) for non normal errors. Some computational and numerical issues are also presented with real data application.