Skip to contents

Changelog

Source: NEWS.md

proxysnr 1.0.1

  • Fix error check in MeanSpectrum() function so that also lists of length 1 can be passed safely to the function.
  • Add additional checks in MeanSpectrum() to cover other possible input errors.

proxysnr 1.0.0

Breaking changes

  • Function names now all follow the recommended style of using verb phrases; hence, the following functions have been deprecated and replaced:

    Calls that use the old name will still work, as they are piped to the respective function of the new name but with a deprecation warning being issued.

  • Handling of transfer function data (#12):

    • SeparateSignalFromNoise(), WrapSpectralResults(): the input transfer functions now have to be supplied as spectral objects (i.e. lists with a frequency and a spectral value vector of the same length);
    • following this, WrapSpectralResults() no longer offers the possibility of directly supplying the respective correction vector via signalling inverse.tf = TRUE and this parameter has been removed accordingly, and SeparateSignalFromNoise() expects a list input (i.e. the spectral object of the transfer function) and no longer a vector of correction (= inverse transfer function spectral) values;

  • to allow for more flexibility in transfer function input, the transfer function spectral objects are internally interpolated to the target frequency axis of the proxy spectral data, so e.g. the diffusion transfer function can be calculated on a higher temporal resolution and still directly supplied to SeparateSignalFromNoise() without requiring manual interpolation as previously and thus reducing code overhead for the user (see vignette("calculate-transfer-functions"). This internal interpolation of course requires sufficient frequency overlap between transfer function and proxy data, which is checked and an error being issued if it is not the case;

  • CalculateDiffusionTF(), CalculateTimeUncertaintyTF(): per default, both functions no longer return the entire underlying spectral data but only the transfer function spectral object, with the previous default return behaviour available via actively setting the new function parameter verbose.output to TRUE;

  • related function documentation was updated accordingly.

  • Reorganisation of the package datasets:

    The DML and WAIS isotope datasets and the corresponding diffusion and time uncertainty transfer function datasets are now the only external (user-accessible) datasets, since they effectively serve as package showcase datasets. The following previously exported datasets are therefore downgraded to internal datasets, since they are only used in internal functions and vignettes to reproduce publication plots: clim.site.par, diffusion.length and t15.noise, where the latter replaces the previous combination of the t15 dataset and the internal function TrenchNoise(). As a minor change alongside that, the T15 trench data, used to create t15.noise, is now processed directly from the PANGAEA data archive in order to avoid a suggest dependency on the TrenchR package.

  • The input parameter structure to PlotStackCorrelation() has been swapped such that the function can now directly digest the output of ObtainStackCorrelation(), e.g. by using a pipe operator.

New features

  • New feature and function EstimateCI() that implements a parametric bootstrapping procedure to estimate confidence levels for an existing dataset of signal, noise, and SNR spectra (#14).
  • SeparateSignalFromNoise() now includes the new function parameter measurement.noise to pass a measurement noise level for correcting the proxy noise spectrum, either as a single value (white noise) or as a spectral object (#13).
  • Package documentation now includes the topic ?spec.object, explaining the meaning and expected structure of a proxysnr spectral object. This term is now consistently used throughout the package and function documentation (see also #12).
  • The functions LPlot() and LLines() to plot/add a spectrum on a double-logarithmic scale are now available as exported functions.
  • A new vignette("proxysnr") with an introduction to the proxysnr package is now available.
  • A github.io homepage for proxysnr is now available, created via pkgdown. In the process of creating that page, the readme, all vignettes, and all function documentation have been revised to render nicely on the homepage (#16).

Minor updates and changes

  • ObtainArraySpectra() now accepts the input proxy data both as a list or as a data frame.
  • ObtainArraySpectra(), SeparateSignalFromNoise(): the output list of both functions now possesses the new attribute array.par storing information on the original proxy record array (number of proxy records, number of observations (e.g. time steps) and (time) resolution), which makes these information more easily available, especially for their use in EstimateCI() (#15).
  • The magrittr pipe operator is now imported into the proxysnr namespace, since with it several proxysnr functions can be executed consecutively; e.g. ObtainArraySpectra() %>% PlotArraySpectra(), ObtainArraySpectra() %>% SeparateSignalFromNoise(), ObtainStackCorrelation() %>% PlotStackCorrelation() (see also #3).
  • PlotStackCorrelation(): The color palette parameter has now a default value of NULL, signalling to set a default color palette internally (taken from the ColorBrewer 2.0 collection), which makes use of the function easier and quicker.
  • GetIntegratedSNR(), ObtainStackCorrelation(): setting of the integration limits has been simplified with the new options to either supply a lower and upper frequency axis index, or directly a frequency range.
  • PlotTF(): It is now possible to only plot a single transfer function type, e.g. only a diffusion transfer function or only a time uncertainty transfer function (#5).
  • CalculateDiffusionTF(), CalculateTimeUncertaintyTF():
    • the output variables from both functions now include an attribute with information on the transfer function simulation run;
    • log-smoothing of the simulated transfer functions can now be switched on via the new function parameter df.log.
  • PlotArraySpectra(): Removal of a certain number of spectral estimates on the low and/or high-frequency side is no longer hard-coded internally in the function but now settable via the new function parameter remove, with its default setting reproducing the previous default behaviour.
  • The internal function SetPlotPar() to set default plotting parameters has been classified as obsolete and removed.
  • Changing the default plotting parameter for the internal plotting functions that create the paper plots has been simplified.
  • PublicationSNR(): Further modified data input arrangement to make function work better in conjunction with WrapSpectralResults() and to allow selection of the data correction version (new function parameter data).
  • Various small bug fixes.

Technical updates

  • Availability of the suggested package simproxyage is now properly being checked at the relevant places in the code, and detailed user information how to install the package are issued if it is not yet installed.
  • The random seed in the vignette on time uncertainty transfer function calculation is now correctly set to ensure reproducibe results in every dataset case.
  • Unit tests have been added, including an automatic GitHub coverage workflow (current code coverage > 97%) (#4).
  • The README now includes tags for GitHub and Zenodo package versions, the R CMD check status and the code coverage, and it has been updated to reflect the latest status of package availability and documentation.
  • Source code now uses 2-space indentation throughout, and all roxygen tags now use only the #' flag.
  • Update to the raw WAIS data processing: the previous automatic download of the original published WAIS data has broken due to changes on the data repository side. Therefore, to ensure reproducibility, the corresponding data has now been manually downloaded and placed into the data-raw folder, from where it is processed into the package dataset, which exactly reproduces the already existing exported dataset in the variable wais. For this reason, and to preserve the original processing date attribute which reflects the time of the corresponding paper publication, the wais dataset variable was not recreated.
  • Vignettes are no longer stored as rendered files under doc/ since that is not supported by R CMD CHECK.
  • Most utility and paper plotting functions are no longer documented.
  • Package license and contributor information has been updated.

proxysnr 0.2.3

#257bddb * WrapSpectralResults(): output element f.cutoff with the cutoff frequency for constraining the diffusion correction is now found within each output element of the raw or corrected spectra, but it has value NA for the elements where diffusion has not been corrected for. * Changed data input arrangement of internal function PublicationSNR().

proxysnr 0.2.2

#fa9171b * DiffusionTF(): Added new function parameter window for analysing only a subset of the total simulated record length. Function documentation updated.

proxysnr 0.2.1

#17d5ecb * DiffusionTF(): Bug fix in the error check on correct input vector length.

proxysnr 0.2.0 (#1)

New function and user-relevant changes

  • The function IntegratedSNR() has been added, which takes the job of calculating the signal-to-noise ratio from the integrated signal and noise spectra, so that it delivers the SNR corresponding to varying temporal record resolutions (compared to the SNR at a specific frequency obtained from SeparateSpectra(), which corresponds to a band-pass filtered record). StackCorrelation() internally now uses the new function for an improved code structure.
  • Parameter N to StackCorrelation() no longer indicates the maximum number of records but is treated directly as a vector of record numbers; so any call which previously used N = 5 now needs to use N = 1 : 5.
  • The rather verbose output of StackCorrelation() has been reduced to the minimum needed - a list of two elements: the frequency axis and the correlation matrix. Output from StackCorrelation() which is piped to the plotting function PlotStackCorrelation() has to be adapted accordingly.

Minor changes

  • All code and documentation has been overhauled so that R CMD check passes with zero errors, warnings or notes.
  • The location of built vignettes in the source package for documentation on GitHub was changed to reflect recent devtools usage.
  • The reference to Muench and Laepple (2018) was updated to reflect the final publication.

proxysnr 0.1.0

  • Package version released along with the publication Münch, T. and Laepple, T.: What climate signal is contained in decadal to centennial scale isotope variations from Antarctic ice cores?, Clim. Past, 2018.