Skip to content

NEWS.md

Latest commit

 

History

History
148 lines (97 loc) · 5.74 KB

NEWS.md

File metadata and controls

148 lines (97 loc) · 5.74 KB

CondCopulas 0.2.0

New features

  • New function measures_nonsimplifyingness_NP for the estimation of the non-parametric measures of non-simplifyingness for conditional copulas, introduced in arXiv:2504.07704.

  • Add more explicit warnings in estimateNPCondCopula when the bandwidth h is too small. This should reduce the total number of warnings there when using too small bandwidth. The previous behavior was the following: there would be 1 warning per value of X3 such that the total sum of kernel-based weights is 0. The new behavior is the following: there is 1 warning per call of estimateNPCondCopula.

  • Inputs Z and newZ to CKT.kernel can be of class data.frame, provided that all the entries are numeric (and that the dimensions match).

  • Add option warnNA = FALSE to CKT.kernel for not raising warnings if some of the estimated conditional Kendall's tau are NAs. The corresponding warning has also been improved and is more explicit about the number of NAs produced, in particular, if the input already contained NAs.

  • CKT.kernel now returns an S3 object of class estimated_CKT_kernel with se, confint and plot methods.

  • In case of cross-validation, CKT.kernel also returns the result of the corresponding function (CKT.hCV.l1out or CKT.hCV.Kfolds).

  • New option nPairs = "all" in CKT.hCV.l1out to use all possible pairs. Also available in CKT.kernel if CKT.hCV.l1out is chosen as the cross-validation method.

  • New S3 generic se for the standard errors. Currently only used for objects returned by simpA.kendallReg and CKT.kernel.

Bug fixes

  • In simpA.NP and simpA.param, the parameter truncVal used to have no effect (contrary to what was written in the documentation). It is now taken into account. Also it is now checked that this parameter belongs to the interval [0, 1/2) so that the integration interval is not empty.

  • In simpA.NP and simpA.param, the weights involved in the Gaussian quadrature have been fixed to correspond to the correct integration interval.

  • Fix a bug in simpA.NP for computation of the integration-based test statistics T1_CvM_Cs3, T1_CvM_Cs4, tilde_T0_CvM, which were incorrect of a constant factor. P-values were unaffected by this bug, as the constant factor was also present in the bootstrap test statistic.

  • Some errors occurring in CKT.kernel with a multivariate Z and small bandwidth have been fixed.

  • Fix bug in computation of the standard errors in simpA.kendallReg. This affected also the corresponding p-values of each estimated coefficients (but not the p-value of the Wald-type test which are computed in another way).

CondCopulas 0.1.4.1

CondCopulas 0.1.4

  • Argument names have been shortened: functions taking arguments of the form observedX1 = are now simplified to X1 = ... (which was already an interface present for some of the functions of the package).

  • Errors because of inputs of different lengths (for example computing some kind of dependence between two vectors X1 and X2 of different lengths) are now of the class DifferentLengthsError. Their messages now explicitly give the lengths of the different objects.

  • Function CKT.kernel can now handle arguments X1, X2 and Z in the case where these are matrices with 1 column.

  • Function simpA.kendallReg can handle the case where only one regressor is given. It also uses stats::lm.fit for the unpenalized regression. A typo in the Wald test statistic has been fixed. Its output is an S3 object of class simpA_kendallReg_test with print, plot, coef, and vcov methods.

  • Function CKT.kernel has now more options to control the possible display of the progress bar to show the progress of the computation.

  • New dependency: testthat has been added to Suggests.

  • New dependency: DiagrammeR has been added to Suggests (for nice plotting of the tree generated by bCond.treeCKT). This package was already suggested by data.tree (which is imported).

  • Fix CRAN NOTE:

    • Found the following Rd file(s) with Rd \link{} targets missing package anchors:

      estimateParCondCopula.Rd: VineCopula

      Please provide package anchors for all Rd \link{} targets not in the package itself and the base packages.

CondCopulas 0.1.3

  • Adding a warning to CKT.kernel() when some estimated conditional Kendall's taus are NA because of a too small bandwidth.

  • Fixing a bug for CKT.kernel() when the conditioning variable is multivariate.

  • Adding and updating references for conditional copulas with discretized conditioning events.

  • Fix an error when running bCond.simpA.CKT().

  • Fix default value of the argument minSize in bCond.treeCKT() to be minSize = minProb * nrow(XI) as intended.

  • Functions CKT.kernel() and CKT.estimate() now warn and return numeric(0) when the argument newZ is numeric(0).

CondCopulas 0.1.2

  • New dependence wdm instead of pcaPP for fast computation of Kendall's tau.

  • Functions CKT.kernel, CKT.hCV.l1out, CKT.hCV.Kfolds and CKTmatrix.kernel gain a new choice "wdm" for the argument typeEstCKT. This new choice allows for faster computation using the package wdm. For observations without ties, it is equivalent to the previous choice typeEstCKT = 4.

CondCopulas 0.1.1

  • Adding DOIs in the DESCRIPTION and small documentation fixes.

CondCopulas 0.1.0

  • Initial release