******************************************
*** Changes for R-package plsdof ***
*** Nicole Kraemer ***
******************************************
--- Version 0.2-6 (March 19, 2013) ---
- Cross-validation is now based on mean squared error as well as on the correlation to the response.
- Instead of cross-validation based on a random split, the split can also be specified using the groups option in the
functions pls.cv, pcr.cv, and ridge.cv.
- You can specify if you want to compute the jackknife coefficients. For very high-dimensional data sets, we recommend
to set conpute.jackknife to FALSE.
- Supervised PCR: You can choose if you want to sort the principal components in pcr and pcr.cv according to their squared correlation
to the response.
--- Version 0.2-5 (February 06, 2013) ---
- The runtime for pcr and pcr.cv is improved, especially in the p>>n scenario.
(1) Instead of an eigen decomposition of t(X)%*%X, a singular value decomposition of X is used.
(2) It is possible to specify the maximum number of principal components