The huge amount of data supplied by advanced technologies, requires a much needed treatment by new processes. So, an improved usage of classic methods, already learned in previous studies, is to be encouraged. When the investigated phenomenon is to be considered from a number of different viewpoints, it is hard the proper choice of a method for analysis. The said choice is linked to the scope of analysis: the multivariate analysis is the collective name for a melting pot of techniques: reduction and organization of the data, search of dependence among variables, statistical inferences. The course is intended a san offer of general ideas and methods at practicable level for people already in possess of basic knowledge of Statistics. Also, the course hopes to make immediately usable the explained methods as soon as possible.

Some basic knowledge of elementary statistics is suitable, however, the lessons are open to everybody.

Aspects of Multivariate Analysis and applications of Multivariate Techniques. Examples in the field of Engineering Natural Sciences and Human Sciences The organization and representation of the data. The geometry of the sample. The multivariate normal distribution. Inferences about a mean vector. Hotelling’s T2 test. Confidence regions and simultaneous comparisons of components means. Comparing mean vectors from two populations. Comparing mean vectors from several multivariate populations: Multivariate Analysis of Variance (MANOVA methods: one way, two way, one way with interaction, ...). Multivariate Analysis of Variance-Covariance (MANCOVA). The classic solution: multiple regression model. Inferences about the regression model. Multivariate multiple regression Data reduction and interpretation: Principal Components Analysis and Factor Analysis. Comparing Principal Components Analysis and Ridge Regression, Partial Least Square and Total Least Square. Comparing Factor Analysis and Principal Component Analysis. Discrimination and classification: Fisher’s discriminant function to separate two or several populations. Comparison with Logistic regression. The program will take place in 15 hours. A final seminar of 5 hours on aspects of time series and Krieging is foreseen (care of prof. Manzino).

On site

Oral presentation

P.D.1-1 - November