maltekurz / pacotest Goto Github PK

View Code? Open in Web Editor NEW

3.0 2.0 2.0 529 KB

R-package: Testing for Partial Copulas and the Simplifying Assumption in Vine Copulas

License: Other

C++ 36.23% R 61.80% C 1.97%

r copula data-science statistics cran

pacotest's People

Contributors

Stargazers

Watchers

Forkers

guhjy swapank

pacotest's Issues

floor call with integers is ambigious

solaris compiler complains about overloaded call to function floor().
Just like #14

Fully integrate the asympt with ranks option in pacotest and pacotestset

value nan is outside the range of representable values of type 'unsigned int'

Check especially file Grouping.cpp

Prevent assigning nan's to int

Causes memtest error's under clang compiler. Just like, #15

Replace stop message by warning in pacotestRvineSeq

pacotestRvineSeq efficient computation of CPITdata

pacotestRvineSeq could be made faster if one implements a subtree extraction for CPITdata instead of computing CPITdata again and again.....

check default value for AggInfo

Change naming from ECORR to CCC

Make the ECORR naming still applicable as depreciated functionality but implement a redirection to the CCC naming convention

Initilization of arma::mat and arma::umat for exporting the decision trees

The initialization of the variables SplitQuantile, SplitThreshold and SplitVariable doesn't work appropriately in C++ and armaldillo, if one just set them to zero. Use .zeros() instead.

Seems to be partly related to #2.

Installation failed on solaris

Error message:

call of overloaded ‘pow(int&, int)’ is ambiguous

Add windows build status via appveyor

R-universe builds failed

See https://github.com/r-universe/maltekurz/actions/runs/1015217090

Export of decision trees to R returns wrong results

The export of decision trees from C++ into R returns wrong results in the case of AggPvalsNumbRep>1. The list of decision trees is filled by replications of the first decision tree but not by all the decision trees used for the aggregation approach.

In the Matlab interface the bug doesn't seem to be present.

Prevent bound violations in `hfunDerivCpit1()` etc

Set direction for numerically computing the gradient.

Omega matrix for asympt with ranks

So far only the omega matrix for the vine copula parameters are correctly handled in the case of asympt with ranks. What's missing are the omega matrix entries for the estimating equations for the correlations in groups and the corresponding cross terms with the vine copula parameter estimating equations.

Failing builds on github actions with macOS

See https://github.com/MalteKurz/pacotest/runs/1720529211
Difficult to track down what caused it, but it might be related to this change in r-lib/actions: r-lib/actions#229