pqrBayes: Bayesian Penalized Quantile Regression
The quantile varying coefficient model is robust to data heterogeneity,
outliers and heavy-tailed distributions in the response variable due to the check
loss function in quantile regression. In addition, it can flexibly model the dynamic
pattern of regression coefficients through nonparametric varying coefficient
functions. Although high dimensional quantile varying coefficient model has been
examined extensively in the frequentist framework, the corresponding Bayesian variable
selection methods have rarely been developed. In this package, we have implemented
the Gibbs samplers of the penalized Bayesian quantile varying coefficient model with
the spike-and-slab priors [Zhou et al.(2023)]<doi:10.1016/j.csda.2023.107808>.
The Markov Chain Monte Carlo (MCMC) algorithms of the proposed
and alternative models can be efficiently performed by using the package.
Documentation:
Downloads:
Linking:
Please use the canonical form
https://CRAN.R-project.org/package=pqrBayes
to link to this page.