Advanced Predictive Models

(DSC 383)

Request Info

Advanced Predictive Models for Complex Data covers random/mixed effects models for multilevel data (clustered data, repeated measures, and longitudinal data) and Gaussian process models for dependent data.  Emphasis is placed on both interpretation of inferences on model parameters and prediction.  The second part of the course introduces nonparametric regression models, including kernel regression, additive models, and random forests.  The course introduces resampling, including the bootstrap, as a tool for quantifying uncertainty in nonparametric regression functions.  A primary goal of the course is for students to be able to select and successfully apply appropriate advanced regression models in applied settings.  The use of statistical software (R) for model fitting, evaluation, and selection is emphasized.

What You Will Learn

  • How to identify different types of dependencies in outcome variables
  • How to select appropriate advanced regression models in applied settings
  • The benefits/limitations of parametric and nonparametric methods
  • How to quantify uncertainty in predictions
  • How to assess model fit

Syllabus

  • Review of the (generalized) linear regression model
  • Multilevel data structures and random effects models
  • Regression models for dependent outcomes
  • Gaussian processes and the spatial (generalized) linear mixed model
  • Nonparametric regression models, including kernel regression, additive models, and random forests
  • Resampling methods and the Bootstrap
  • Evaluating model fit and model selection

Estimated Effort

10-12 Hours/week

Course Availability

  • Spring 2022

Take the Next Step

Advance your career with UT Austin's Master of Data Science Online.