Advanced Data Analytics: Statistical Learning and Latent Variables


This course combines the perspectives of latent variable modeling and statistical learning to uncover relationships and processes within a given data set. Topics covered will include principal component analysis and factor analysis and extensions (mixtures of factor analyzers, factor mixture models, etc.), latent trait models, latent class models, and mixture models. Connections will be made between latent variables and modern statistical learning methods (both supervised and unsupervised), with the associated topics of support vector machines, additive trees, and other advanced classification schemes being covered for supervised learning. Unsupervised learning will include a discussion of a variety of methods for finding "groups" within the data.

From a practical point of view, these methods will be taught and operationalized through the analysis of real data in a true "workshop"-like format in which hands-on skills are emphasized. Attention will be paid to how to best interpret and present results in presentations or publications. Some examples of practical decision points that will be discussed include: (1) how to select variables, (2) how to standardize/transform variables, (3) which approaches are appropriate for which data (for example, cross-sectional vs. longitudinal data), (4) how to reduce dimensionality, and (5) how to determine if your final solution is both valid and generalizable.

Fee: Members = $1500; Non-members = $3000

Tags: Data Analytics, Statistical Learning, Latent Variables

Course Sections