The goal of this problem is to demonstrate the use of principal components analysis in the selection of a minimal informative set of explanatory variables. We use data sets MORN.mat and AFT.mat used in the text. Open and run the (data) script file indicators.asc to create these two data sets.
1. Use the kernel method to regress the afternoon volatility ratio (third column of the data matrix SAFT.mat) against the six variables of SMORN.mat, and compute the sum of square errors. (NB: explain your bandwidth choice).
2. Perform a principal component analysis of the SMORN data set and compute the vector of the daily values of the two most important components PC1 and PC2. Use the kernel method to regress the afternoon volatility ratio against PC1 and PC2, and compute the sum of square errors. Again, you will need to justify your choice of bandwidth. Compare with the sum of square errors found in question 1, the sum of square errors found in the experiment with the kernel method described in the text. Comment.