D-Vine Quantile Regression

Implements D-vine quantile regression models with parametric or nonparametric pair-copulas. See Kraus and Czado (2017) and Schallhorn et al. (2017) .


Travis-CI Build Status AppVeyor Build Status Coverage status CRAN status

An R package for D-vine copula based mean and quantile regression.

How to install

  • the stable release from CRAN:

    install.packages("vinereg")
  • the latest development version:

    devtools::install_github("tnagler/vinereg", build_vignettes = TRUE)

Functionality

See the package website.

Example

set.seed(5)
library(vinereg)
data(mtcars)
 
# declare factors and discrete variables
for (var in c("cyl", "vs", "gear", "carb"))
    mtcars[[var]] <- as.ordered(mtcars[[var]])
mtcars[["am"]] <- as.factor(mtcars[["am"]])
 
# fit model
(fit <- vinereg(mpg ~ ., data = mtcars))
#> D-vine regression model: mpg | disp, hp, gear, carb, cyl, am.1, wt, vs, qsec, drat 
#> nobs = 32, edf = 95, cll = -56.09, caic = 302.19, cbic = 441.43
 
summary(fit)
#>     var     edf          cll       caic       cbic      p_value
#> 1   mpg 3.2e-11 -98.59271949 197.185439 197.185439           NA
#> 2  disp 2.0e+00  29.53428159 -55.068563 -52.137091 1.490817e-13
#> 3    hp 3.0e+00   2.33231128   1.335377   5.732585 1.980680e-01
#> 4  gear 5.0e+00   2.16379041   5.672419  13.001099 5.032784e-01
#> 5  carb 7.0e+00   2.25663902   9.486722  19.746873 7.191178e-01
#> 6   cyl 9.0e+00   1.57187334  14.856253  28.047876 9.583219e-01
#> 7  am.1 1.0e+01   1.75059329  16.498813  31.156172 9.670581e-01
#> 8    wt 1.3e+01   1.62623809  22.747524  41.802091 9.968597e-01
#> 9    vs 1.3e+01   0.52958111  24.940838  43.995405 9.999946e-01
#> 10 qsec 1.6e+01   0.70430954  30.591381  54.043155 9.999992e-01
#> 11 drat 1.7e+01   0.02886845  33.942263  58.859773 1.000000e+00
 
# show marginal effects for all selected variables
plot_effects(fit)
#> `geom_smooth()` using method = 'loess' and formula 'y ~ x'
 
# predict mean and median
head(predict(fit, mtcars, alpha = c(NA, 0.5)), 4)
#>       mean      0.5
#> 1 19.34836 19.36129
#> 2 19.17641 19.19810
#> 3 25.28064 25.13942
#> 4 19.70841 19.67779

Vignettes

For more examples, have a look at the vignettes with

vignette("abalone-example", package = "vinereg")
vignette("bike-rental", package = "vinereg")

References

Kraus and Czado (2017). D-vine copula based quantile regression. Computational Statistics & Data Analysis, 110, 1-18. link, preprint

Schallhorn, N., Kraus, D., Nagler, T., Czado, C. (2017). D-vine quantile regression with discrete variables. Working paper, preprint.

News

vinereg 0.5.0

DEPENDS

  • require rvinecopulib (>= 0.3.0) due to breaking changes in this package.

BUG FIXES

  • prevent nan errors in loglik calculation.

  • allow for empty and bivariate models.

  • properly pass degree parameter for margin al estimation.

vinereg 0.4.0

BUG FIXES

  • Fix handling of uscale in fitted.vinereg().

  • Fix handling of mult parameter for pair-copula fits in vinereg().

  • Fix orientation of asymmetric pair-copulas.

vinereg 0.3.0

DEPENDS

  • Use furrr and fututre packages instead of parallel, doParallel, and foreach for parallelization.

NEW FEATURES

  • New print() and summary() generics for vinereg objects.

  • New plot_effects() method to show the marginal effects of variables.

  • Allow to predict the mean with predict(object, alpha = NA).

vinereg 0.2.0

  • First official release.

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("vinereg")

0.7.4 by Thomas Nagler, 5 months ago


https://tnagler.github.io/vinereg/


Report a bug at https://github.com/tnagler/vinereg/issues


Browse source code at https://github.com/cran/vinereg


Authors: Thomas Nagler [aut, cre] , Dani Kraus [ctb]


Documentation:   PDF Manual  


GPL-3 license


Imports rvinecopulib, kde1d, Rcpp, assertthat

Suggests knitr, rmarkdown, ggplot2, AppliedPredictiveModeling, quantreg, tidyr, dplyr, purrr, scales, mgcv, testthat, covr

Linking to rvinecopulib, RcppEigen, Rcpp, BH, wdm, RcppThread, kde1d


See at CRAN