R: News for package 'psd': CRAN (original) (raw)

Fixed an issue related to how psd initiates its internal options.
New flag in psdcore indivating if calculations are multivariate or not; this will be used in the future to make plotting functions more robust.
With the addition of a new function na_locf, the dependency on zoo has been removed.

Major bump which adds multivariate calculations (i.e., the cross spectrum) using the same optimization feature.

Significant speed improvements in spectral derivative computation implemented in riedsid2thanks to @jkennel; this will be most apparent for very long timeseries.
riedsid is deprecated
niter in pspectrum default set down from 5 to 3
Modernized revdep checks

Addresses three minor c++ issues associated with type overloads (not seen until CRAN-build-checks):
1. resample_fft.cpp:92:22
2. resample_fft.cpp:104:31
3. resample_fft.cpp:264:34
Prevents test-failure when fftw is not available. Thanks again, fftw!

This is a major version bump that addresses performance issues by converting known bottlenecks to c++ though the Rcpp package. There has also been quite a lot of cleanup and rearranging of documentation, and making sure methods are consistent. Some attempts to be backwards compatible have been made, but don't be surprised if results from previous versions are different – sorry!
psdcore now uses resample_fft_rcpp which is a lightning-fast (by comparison with pure-R) implementation of the fft-resampling/reweighting method we use; this new function leverages the power of RcppArmadillo.
ctap_simple has been superceded by ctap_simple_rcpp; in the process of rewriting in c++, a minor bug in the c implementation was found (and fixed).
Issues/pull-requests can be entered on the Issues page.
Taper constraints: ctap_simple_rcpp/ctap_simple and ctap_loess gain their own man-page; ctap_markov and ctap_friedman were previously made defunct, but are now permanently removed from the code-base.

Thanks to David Myer for catching a number of mistakes in the code and documentation, including a bug in riedsid from an improper choice of logarithm base.
Removed the Local.loss argument from riedsid

We have published a paper in Computers & Geosciences regardingpsd, and the citation is up to date: citation('psd')If you cannot access the publication (10.1016/j.cageo.2013.09.015), contact Andy for a reprint.
Fixed an example which was causing CRAN-buildcheck errors.
Fixed a small error in the normalization vignette – thanks to Richard Gaal for spotting this.

The taper-constraint methods ctap_markov and ctap_friedman are now defunct.
Removed 'units.sty' dependency in vignettes (was causing NOTES and unhappiness).
VignetteEngine added, and :::'s removed, in order to ensure compatibility with R-devel.
Added citation information for upcoming paper in Computers and Geoscience.

Revised discussions regarding AR response spectrum.
Fixed psdcore: it was applying parabolic weights incorrectly.
Changed argument name ntap_pilot in pspectrum to ntap.init.
Changed ceiling in minspan to round.
Fixed km field in magnet dataset.
Fixed referencing s.t. ?plot isn't ambiguous.
The default number of iterations and pilot tapers was changed to 3 and 7 respectively.

This marks the first release to CRAN: psd. We have included three vignettes, one of which is an overview of the general functionality of the program. These may be listed withvignette(pack="psd"), but we recommend consultingvignette("psd_overview", package="psd") to begin.
Altered the environment manipulation features to prevent .GlobalEnv modification.
Fixed CRAN-check failure with usepackage{color} in vignettes.
Removed empty .Last.lib function to prevent CRAN-check note.