Why are some algorithm options (SVD, Kernel) disabled for PCA, PCR and PLSR?


Please explain why do you think this question should be reported?

Report Cancel

This is usually because there are missing values in the selected data. While the NIPALS algorithm handles this by imputing the missing values automatically, the SVD and Kernel-options require non-missing data. Missing values can be imputed using Tasks – Transform – Fill Missing before modelling (although this is probably less efficient than just running the NIPALS algorithm directly).

Note that in the special case where categorical variables are included as predictors (which is allowed if they contain 2 levels only), non-assigned levels will not be recognized as missing, and SVD and Kernel-options will be available. An attempt to perform calibration without switching to NIPALS in this case will result in a low-rank warning and no model will be returned.

About the Author

Leave an answer