Multivariate Analysis

This course provides a practical introduction to the analysis of high dimensional data, consisting of correlated variables. Objectives may include data reduction, summarisation, searching for groupings and classification, which are common in the areas of unsupervised and supervised learning. The course sequentially covers a set of classic multivariate analysis techniques for meeting these objectives, starting with principal components analysis and finishing with partial least squares. R, SAS, SPSS or Stata may be used for practical work.

Course Code FSSS3RIMA
Course Leader Dankmar Böhning

Course Description
Overview

Multivariate Analysis is concerned with methods of analysing data that consist of observations on two or more variables for each individual or unit. Multivariate data will generally be correlated, and a wide variety of techniques are available to analyse these data. Many scientific disciplines make use of such techniques for investigating and understanding large datasets.

During this course, commonly used multivariate techniques are introduced, developed, and relationships between them examined. The emphasis of the course is practical application and interpretation of results using a range of scientific related examples. Mathematical details are kept to a minimum.

All practical work may be done using R, SAS, SPSS or Stata. Each participant may choose one of these statistical packages, according to their preference, and conduct all practical exercises using that package. Please note SPSS and Stata users will need to use the freely available R software for practical work on partial least squares. Examples used will be drawn from a variety of applications in science and life science.

Who Should Attend?

Scientists and related involved in analysing multivariate data. A working knowledge of basic statistics and linear regression will be assumed, such as standard deviations, correlations, simple and multiple linear regression

How You Will Benefit?

You will gain a sound practical understanding of commonly used multivariate techniques.

What Do We Cover?

• Principal components analysis (PCA)
• Factor analysis and the use of rotations
• Cluster analysis for grouping objects
• Multidimensional scaling for mapping and relationship to PCA
• Canonical variates analysis for highlighting differences between groups
• Discriminant analysis and allocation rules
• Correspondence analysis
• Partial least squares regression

Choice of Software

Practical work may be done in one of the following statistical packages for the duration of the course: R, SPSS, SAS or Stata. The exceptions are SPSS and Stata, whose users will need to use the freely available R software for practical work on partial least squares.