fastpns carried out pns but also has an option to first reduce dimension by PCA to n.pc, then fit pns. Useful for high-dimensional data, e.g. dimension > 30.
Fixed a bug in the plot of the PNS mean in the function pns
New backfit function for backfitting from PNSS or PCA scores
Added tangentcoords option to shapes.cva and output all the CV scores (rather than just 3)
Improved the 3d sphere plot for pns()
speed up pnss3d when n < km-m(m-1)/2-m
Added in pnss3d and plot3darcs for displaying the PNSS modes of variation
Faster versions of some functions kindly supplied by Gregorio Quintana-Orti and Amelia Simo, University Jaume I, Spain.
Corrected bug in estcov for method="Power" when exit occurred in some zero eigenvalue cases, by including abs(eigenvalue)
Added Principal Nested Spheres (pns) Added Principal Nested Shape Spaces using PCA (pns4pc) Updated some references in help files
Minor adjustment to Penalized Euclidean Distance regression function, including a different name ped()
Added in a function to carry out Penalized Euclidean Distance regression, which is a sparse regression method (Vasiliu et al., 2017, arxiv).
Renamed the function sigma() to sigmacov() prevent a warning that the same function name is used in the `stats' package.
corrected a bug in the calculation of principal warp eigenvectors in the function shaperw, which in turn is used by procGPA (thanks to Paolo Piras)
corrected an error in apes$x[,,60] data, which should have been the same as panf.dat[,,1] (thanks to Katie Severn)
Corrected a bug in shaperw for the m=3 case (transposes needed) (thanks to Valerio Varano and Paulo Piras)
internal expression of bendingenergy (benergy in TPSgrid) has correct constant now. (thanks to Valerio Varano)
procdist - function added to compute different types of Procrustes shape distances
MDSshape - function added to compute MDS mean shape
Several new datasets added
procGPA fixed recently introduced error in reading in complex matrices
procGPA( , scale=FALSE,pcaoutput=FALSE) was still calculating PCA, so this has now been fixed.
The internal function prcomp1 now uses eigen() rather than svd(), due to some convergence issues in LAPACK for some singular matrices.
transformations() :relative translations between centroids now given, rather than just translating the original to have centroid at the origin.