Estimation of Bayesian vectorautoregressions with/without stochastic volatility.
Implements several modern hierarchical shrinkage priors, amongst them Dirichlet-Laplace prior (DL), hierarchical Minnesota prior (HM), Horseshoe prior (HS), normal-gamma prior (NG), $R^2$ -induced-Dirichlet-decomposition prior (R2D2) and stochastic search variable selection prior (SSVS).
Concerning the error-term, the user can either specify an order-invariant factor structure or an order-variant cholesky structure.
Install CRAN version:
install.packages("bayesianVARs")
Install latest development version directly from GitHub:
devtools::install_github("luisgruber/bayesianVARs")
The main workhorse to conduct Bayesian inference for vectorautoregression models in this package is the function bvar().
Some features:
predict(), plot(), coef(), vcov() and fitted().specify_prior_phi() and specify_prior_sigma().set.seed(537)
# load package
library(bayesianVARs)
# Load data
train_data <-100 * usmacro_growth[1:237,c("GDPC1", "PCECC96", "GPDIC1", "AWHMAN", "GDPCTPI", "CES2000000008x", "FEDFUNDS", "GS10", "EXUSUKx", "S&P 500")]
test_data <-100 * usmacro_growth[238:241,c("GDPC1", "PCECC96", "GPDIC1", "AWHMAN", "GDPCTPI", "CES2000000008x", "FEDFUNDS", "GS10", "EXUSUKx", "S&P 500")]
# Estimate model using default prior settings
mod <- bvar(train_data, lags = 2L, draws = 2000, burnin = 1000, sv_keep = "all")
# Out of sample prediction and log-predictive-likelihood evaluation
pred <- predict(mod, ahead = 1:4, LPL = TRUE, Y_obs = test_data)
# Visualize in-sample fit plus out-of-sample prediction intervals
plot(mod, predictions = pred)
bayesianVARs - Shrinkage Priors for Bayesian Vectorautoregressions in R
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4