Content, progress and pedagogy of the module

This course is about how to extract and interpret useful information from experimental and process data, including large datasets with thousands of samples and variables (e.g. spectroscopic data). It covers all main steps of data analysis, including exploratory analysis, multivariate regression and validation, classification, preprocessing and variable selection.

Learning objectives

Knowledge

Students who complete the module must be able to

• account for general methods for multivariate data analysis (principal component analysis, multiple linear regression, principal component regression, projection on latent structures, soft independent modelling of class analogy)
• account for methods for data preprocessing (centering, scaling, nonlinear and spectroscopic reprocessing, orthogonal signal correction).
• explain basic methods for variable selection (Selectivity ratio, VIP, interval PLS, jack-knife)
• explain the theoretical background of these methods, their advantages and limitations as well as possible applications
• explain how multivariate methods complement traditional statistical methods

Skills

• explore multivariate data, find groups and trends, detect and remove outliers
• calibrate and do proper validation of multivariate regression models, use these models for prediction
• evaluate if data need a preprocessing and which method to apply
• calibrate and evaluate models for data classification
• compare different regression and classification models and identify the best
• use multivariate methods for analysis of real data from different applications

Type of instruction

• Lectures
• Workshops
• Exercises
• Mini-projects

Extent and expected workload

150 hours

Exam

Exams