Summary: This course teaches you techniques for fitting statistical models to identify important variables. Manual, graphical, and automated variable selection techniques are presented, along with advanced modeling methods. The demonstrations include modeling both designed and undesigned data. Techniques are illustrated using both JMP software and JMP Pro software. Note that JMP Pro software is needed for the advanced techniques covered in the second half of this course.
Duration: 14 hours of content.
Modalities:
- live online with instructor -- This course is available periodically in our public course schedule. The public courses are an opportunity to learn this content with a live instructor, but they are currently only offered in English and at times most convenient to a US audience (because most of our instructors are in US time zones). Don't see what you are looking for? Let us know.
- through a third-party training vendor -- Any course in our JMP Curriculum could be taught by a licensed training vendor, including through the training department at your own company. Contact your JMP representative to learn more.
Prerequisites: Before attending this course, it is recommended that you complete the JMP® Software: A Case Study Approach to Data Exploration and JMP®: Statistical Decisions Using ANOVA and Regression courses or have equivalent experience.
Learning Objectives:
- Identify a subset of predictors as important using a statistical model.
- Validate statistical models using cross-validation, holdback validation, and information-theoretic criteria.
- Perform stepwise and all subsets regression.
- Select important predictors using graphical methods and decision trees.
- Perform penalized regression for Gaussian and non-Gaussian responses.
- Use the Generalized Regression platform to identify important predictors.
Course Outline:
Introduction to Variable Selection
- Introduction to models.
- Model validation methods.
- Variable importance.
- OLS regression and backward selection.
Classic Methods of Variable Selection
- Graph-based methods.
- Stepwise regression.
- All subsets regression.
Generalized Regression (JMP Pro Required)
- Introduction.
- Lasso/ridge/elastic net.
- Adaptive lasso/adaptive elastic net.
- Methods with multiple passes.
Special Situations for Generalized Regression (JMP Pro Required)
- Binomial response.
- Poisson response.
- Multiple responses.