cvq2.setB: Small data set to demonstrate the difference between the conventional and the predictive squared correlation coefficient while performing a cross-validation.
Description
Contains a small data set with six observations, the observed value y depends on the parameter $x$.source
Generic data set, created for this purpose only.Details
The prediction power can be determined with cross-validation.
The cross-validation can be performed as Leave-One-Out ($\code{nFold} = \var{N} = 6$) or as k-fold ($\code{nFold} = {2, 3}$).
If $\code{nFold} = {2, 3}$), modelData is randomly split into nFold disjunct and equal sized (test) sets. Furthermore, in that case one has the opportunity to repeat the cross-validation, while each run ($\code{nRun} = 2 \ldots x$) has an individual test set compilation.
The prediction power, $q^2_{cv}$, calculated for this data set is considerably smaller than the model calibration, $r^2$, promises.