This presentation is used for 15 minutes talk at SSC16, Oslo conference. This will try to summarize the prediction and estimation comparison through simulation using the tool called simrel
which is in the form of R-package. The comparison is made on the basis of prediction and estimation error to compare PCR, PLS and Envelope methods. The data is simulated from a specifically designed models where certain aspects such as correlation between predictors, correlation between response, position of relevant predictor components for response are considered. An full factor factorial design is used to to test the effect of each levels of these factor on the methods and their performance on prediction and estimation.
For more details see: a) Prediction Comparison and b) Estimation Comparison
For R-package simrel
used in simulation can be installed from R-package repository CRAN as,
install.packages("simrel")