Compute PCA for bayesLR

[Sum02dean]

This is to overcome issues with insufficient rank and lack of invert-ability of the design matrix using the Bambinos package.

[Sum02dean]

Attempted to use PCA, this threw the predictions way off. The next step is to save the transformer object for each training subset, then when predicting with each estimator, re-transform the test set respectively.

[Sum02dean]

PCA does not seem to work well on this data.