Canonical correlation analysis and maximum correlation via projection pursuit, as well as fast implementations of correlation estimators, with a focus on robust and non-parametric methods.
Changes in ccaPP version 0.3.2
+ Added references for maxCorGrid() and permTest(). + Added vignette 'ccaPP-intro'.
Changes in ccaPP version 0.3.1
+ Added 'diabetes' data from Andrews & Herzberg (1985). + Print method and help file of permutation tests no longer use the term 'independence'. + Functions for robust CCA now standardize dummy variables with mean and standard deviation. + Bugfix in functions for CCA: R session no longer crashes for higher-order canonical correlations in case of high-dimensional data.
Changes in ccaPP version 0.3.0
+ Standardization in functions for CCA and maximum correlation is now optional. + Functions for CCA and maximum correlation now return center and scale estimates from standardization. + Slightly faster version of fastMedian(). + Function permTest() now allows to supply permutations as an argument. + Bugfix in functions for CCA concerning side constraints for higher-order canonical correlations. + Bugfix in functions for CCA: reduced data for higher-order canonical correlations no longer standardized. + Bugfix in corM(): robust starting values now work with dummy variables. + Renamed internal C++ function rank() to fix error with clang++ under OS X Mavericks.
Changes in ccaPP version 0.2.0
+ Functions ccaGrid() and ccaProj() now force consistency of the Spearman, Kendall and quadrant correlation. + New functions maxCorGrid() and maxCorProj() for maximum correlation, which retain the argument for consistency of the Spearman, Kendall and quadrant correlation. + permTest(): now using the new functions maxCorGrid() and maxCorProj() rather than ccaGrid() and ccaProj(). + permTest(): argument 'cl' is now preferred over argument 'ncores' for parallel computing. + Added fallback mode for robust standardization if the MAD of a variable is zero. + Bugfix in ccaGrid() and its wrapper CCAgrid: seed for the random number generator is no longer ignored. + Bugfix in CCAproj(): now correctly wrapped around ccaProj().
Changes in ccaPP version 0.1.1
+ Using package parallel now to detect the number of available processor cores rather than OpenMP. + Internally created parallel cluster in permTest() is now stopped via on.exit(). + Bugfix in permTest(): random number streams for parallel clusters supplied by the user are only reset if a seed is supplied.