1.0.3
- Confidence intervals for the MST provided based on chi-squared
distribution.
1.0.4
- Change nomenclature to omit “the” from \(R^2\).
1.0.6
- Cited publication in The American Statistician rather than
preprint
1.0.7
- Bug fix when supplying vector rather than matrix of regressors
1.0.8
- Multithreading for windows machines in vignette
1.0.9
- Disallow data frames as input
1.0.10
- Set minimum of folds to 3 to prevent problems in CV
1.0.11
- Update Bates and own publication from preprints to published
articles
1.0.12
- Correct BugReports url
- Change Brassica reference to published version
1.0.13
- Switch to repeated CV also for estimating the correlation between
MSE and MST in simpleCV()
- Correct scaling factor for SE of MSE to (nOuterFolds-1)/nOuterFolds
instead of nOuterFolds/(nOuterFolds-1), which was erring on the
conservative side. This error was also present in the code of the
original publication
- Correct bug to propagate … to all predFun and fitFun instances
- Correct bug in fold assignment in estMSE().
- Correct bugs in boostrap MSE and SE estimation.