There is a newer version of this course. Please see the schedule for the new Multivariate Statistics for Understanding Complex Data course.
This course teaches how to apply and interpret a variety of multivariate statistical methods to research and business data. Strong emphasis is on understanding the results of the analysis and presenting your conclusions with graphs.
Learn how to
- make sense of the math behind many multivariate statistical analyses
- reduce dimensionality with principal components analysis
- identify latent variables with exploratory factor analysis and factor rotation
- understand individual preferences with qualitative preference analysis
- explain associations among many categories with correspondence analysis
- finds patterns of association among different sets of continuous variables with canonical correlation analysis
- explain differences among groups in terms of many predictor variables through canonical discriminant analyses
- classify observations into groups with linear and quadratic discriminant analyses
- fit complex multivariate predictive models with partial least squares regression analysis.
Who should attend
Business analysts, social science researchers, marketers, and statisticians who want to use SAS to make sense of highly dimensional multivariate data
Before attending this course, you should be familiar with statistical concepts such as hypothesis testing, linear models, and collinearity concepts in regression. You should have an understanding of the topics taught in Statistics 2: ANOVA and Regression or equivalent.
This course addresses SAS/STAT software.
Overview of Multivariate Methods
Principal Components Analysis using the PRINCOMP procedure
- examples of multivariate analyses
- matrix algebra concepts
Exploratory Factor Analysis using the FACTOR procedure
- principal component analysis for dimension reduction
Multidimensional Preference Analysis using the PRINQUAL and TRANSREG procedures
- factor analysis for latent variable measurement
- factor rotation
Correspondence Analysis using the CORRESP Procedure
- plotting high-dimensional preference data
- mapping preferences to other characteristics
Canonical Variate Analysis using the CANCORR and CANDISC Procedures
- understanding complex associations among categorical variables
Discriminant Function Analysis using the DISCRIM Procedure
- multivariate dimensions reduction for two sets of variables
Partial Least Squares Regression using the PLS Procedure
- classification into groups
- linear discriminant analysis
- quadratic discriminant analysis
- empirical validation
- PLS for one target variable
- PLS for many targets
- PLS for predictive modeling