roahd 1.4.3

New feature

• Added tools for manipulating and visualizing depthgrams (#1, @aefdz). This mathematical constructs aim at facilitating the visualization of outliers in high dimensional functional data sets. The depthgram() function computes a number of depthgrams from the functional data set. An S3 specialized method for plot() makes it possible to visualize the depthgrams and proceed with a visual inspection at outliers.

Minor updates

• Improved unit testing: Refactoring of unit tests using testthat 3e edition and in particular snapshot tests for plots and complex objects.
• Added an hexsticker.

Version 1.4

Changelog

Here's a list of what is changed in this update of roahd:

Upgrades

Major upgrades

1. Extended Spearman's correlation coefficient computation for multivariate datasets with more than two
   components.

2. Added bootstrap-based computation of Spearman's correlation coefficient bias and standard deviation.

3. Added methods to provide bootstrap-based confidence intervals on Spearman's coefficients for two
   univariate functional datasets or a multivairate functional dataset.

4. Added a bootstrap-based test on Spearman's correlation coefficient for two multivariate functional datasets.

5. Added an outliergram version (without graphical display of original data) of multivariate functional datasets.

6. Added example multivariate functional datasets of ECG signals.

Minor updates

1. Added two convenience functions to append compatible functional datasets (univariate or multivariate).

2. Added a [-operator overload for multivariate functional dataset representation mfData.

Fixes

Major fixes

1. Fixed bug in cor_spearman function. Now the standard spearman correlation is not computed on ranks of MHI/MEI, but on
   MHI/MEI itself. The difference is very small, but allows for full reproducibility of the results in the original paper.

Minor fixes

1. Fixed typos in doc

2. Standardised formulas for the application of F inflations in outliergram and boxplot

Version 1.2

Some changes have occurred in this new release, mainly:

1. Removed check for uniformity in the grid of fData() and mfData() constructor

2. Added the possibility to subset fData in time with logical vectors

3. Fixes in methods BD, BD_relative, HI and EI: the previous computational technique was based on arguments from the popular reference "Exact fast computation of band depth for large functional datasets: How quickly can one million curves be ranked?" by Sun, Genton and Nychka, which in the case of BD, and HI/EI are incorrect. Now the implementation sticks to the definition, at the cost of a higher computational burden (and thus, time to complete the computation).

Version 1.1

Here's a list of what is changed in this update of roahd:

Major fixes

1. Modified the check of the grid provided to build fData objects.
   Since support is provided only for evenly spaced grids, a check is needed before building an fData object.
   Before it was:

all( abs( diff( unique( diff( grid  )  )  )  ) < 1e-14  )

Now it is:

max( diff( unique( diff( grid  )  )  )  ) / diff( range( grid  )  ) < 1e-13

which is much more robust in practical cases.

2. Extended README.md

3. Added cov_fun method to compute covariance and cross-covariance functions
   for either univariate or multivariate functional data. Implemented the S3 class
   Cov and plotting specialisation plot.Cov, wrapping graphics::image.

Minor fixes

1. Fixed typos in documentation

2. Fixed typos in vignette

3. Added Travis and Codecov support

4. Modified the default parameter value for trial_size in fbplot from Data$N
   to 8 * Data$N.

5. Added check to fbplot and outliergram that raises warnings when parameters
   different than those supported are provided through adjust argument.