Changes in version 1.2.9                        

In several places use riemdist(x,y,reflect=reflect) so that the reflect
option is correct within functions. Thanks to dhalpern

In pns, pnss3d function changed the default method to BIC (as much
faster)

In pns function changed LR test to use chi-squared_1 critical value

In fastpns included distr= option, as in pns

                 Changes in version 1.2.8 (2025-10-24)                  

Features

pns includes extra options for deciding between great vs small spheres
at each stage, namely ks.test, var.test, distr (LR test using particular
distributions)

fastpns carries 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.

Added an option to pns and pnss3d to choose the mean.type. Previously
was Frechet and now can be Fisher if desired.

Fixes

Fixed a bug in the plot of the PNS mean in the function pns

Included more (1000) values to search for Frechet mean at last stage,
rather than just n (which may be small). Hence this is more accurate.

                 Changes in version 1.2.7 (2023-02-03)                  

Features

  - 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

Fixes

  - changed rgl.open to open3d and rgl.bg to bg3d

                 Changes in version 1.2.6 (2021-03-31)                  

Features

Added in pnss3d and plot3darcs for displaying the PNSS modes of
variation

                 Changes in version 1.2.5 (2019-12-01)                  

Features

Faster versions of some functions kindly supplied by Gregorio
Quintana-Orti and Amelia Simo, University Jaume I, Spain.

Fixes

Corrected bug in estcov for method="Power" when exit occurred in some
zero eigenvalue cases, by including abs(eigenvalue)

                 Changes in version 1.2.4 (2018-08-18)                  

Added Principal Nested Spheres (pns) Added Principal Nested Shape Spaces
using PCA (pns4pc) Updated some references in help files

                 Changes in version 1.2.3 (2017-11-18)                  

Minor adjustment to Penalized Euclidean Distance regression function,
including a different name ped()

                 Changes in version 1.2.2 (2017-11-03)                  

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.

                 Changes in version 1.2.1 (2017-09-22)                  

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)

                 Changes in version 1.2.0 (2017-02-02)                  

corrected an error in apes$x[,,60] data, which should have been the same
as panf.dat[,,1] (thanks to Katie Severn)

                 Changes in version 1.1-13 (2016-07-06)                 

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)

                       Changes in version 1.1-12                        

procdist - function added to compute different types of Procrustes shape
distances

                 Changes in version 1.1-11 (2015-08-18)                 

MDSshape - function added to compute MDS mean shape

Several new datasets added

                 Changes in version 1.1-10 (2014-09-20)                 

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.