See how to use JMP Pro PLS when there are more explanatory variables than observations, highly correlated explanatory variables and responses, and many explanatory variables.
Presenter: Peter Bartell
See how to use JMP Pro PLS when there are more explanatory variables than observations, highly correlated explanatory variables and responses, and many explanatory variables.
Presenter: Peter Bartell
The presenter describes when and why PLS is useful, describes the advantages of the PLS implementation in JMP Pro over that in JMP, and introduces the case studies he will present to demonstrate using PLS in JMP Pro.
The presenter uses consumer ratings for 24 types of bread to demonstrate how to use PLS to identify product attributes to help guide new bread formulation and design processes. He uses leave-one-out cross-validation and shows how to examine and interpret Root Mean Press; NIPALS Fit x and y scores for a single latent factor; Diagnostic Plots; and VIP vs Coefficients Plots. He uses the Prediction Profiler to maximize desirability.
The presenter shows how to create a model to evaluate the levels of three different compounds in spectral emissions of water samples. He uses the Model Comparison summary to see how x and y scores correlate and to identify which variables explain variation. He shows how to save prediction values to the data table, which will be useful to the model if new observations are added.
The presenter uses Discriminant Analysis to demonstrate how to determine if genetic expression information can be used to accurately classify estrogen receptor status. He uses data that includes over 10,000 gene expression characteristics from a study of 230 individuals. He shows how to explore missing values; interpret ROC curves; and examine misclassification, false positives and false negatives.