The main function for estimating a mixed-frequency BVAR.
estimate_mfbvar(mfbvar_prior = NULL, prior, variance = "iw", ...)
a mfbvar_prior
object
either "ss"
(steady-state prior), "ssng"
(hierarchical steady-state prior with normal-gamma shrinkage) or "minn"
(Minnesota prior)
form of the error variance-covariance matrix: "iw"
for the inverse Wishart prior, "diffuse"
for a diffuse prior, "csv"
for common stochastic volatility or "fsv"
for factor stochastic volatility
additional arguments to update_prior
(if mfbvar_prior
is NULL
, the arguments are passed on to set_prior
)
An object of class mfbvar
, mfbvar_<prior>
and mfbvar_<prior>_<variance>
containing posterior quantities as well as the prior object. For all choices of prior
and variance
, the returned object contains:
Array of dynamic coefficient matrices; Pi[,, r]
is the r
th draw
Array of monthly processes; Z[,, r]
is the r
th draw
Array of monthly forecasts; Z_fcst[,, r]
is the r
th forecast. The first n_lags
rows are taken from the data to offer a bridge between observations and forecasts and for computing nowcasts (i.e. with ragged edges).
prior = "ss"
, it also includes:
roots
The maximum eigenvalue of the lag polynomial (if check_roots = TRUE
)
If prior = "ssng"
, it also includes:
psi
Matrix of steady-state parameter vectors; psi[r,]
is the r
th draw
roots
The maximum eigenvalue of the lag polynomial (if check_roots = TRUE
)
lambda_psi
Vector of draws of the global hyperparameter in the normal-Gamma prior
phi_psi
Vector of draws of the auxiliary hyperparameter in the normal-Gamma prior
omega_psi
Matrix of draws of the prior variances of psi; omega_psi[r, ]
is the r
th draw, where diag(omega_psi[r, ])
is used as the prior covariance matrix for psi
variance = "iw"
or variance = "diffuse"
, it also includes:
variance = "csv"
, it also includes:
phi
Vector of AR(1) parameters for the log-volatility regression; phi[r]
is the r
th draw
sigma
Vector of error standard deviations for the log-volatility regression; sigma[r]
is the r
th draw
f
Matrix of log-volatilities; f[r, ]
is the r
th draw
If variance = "fsv"
, it also includes:
latent
Array of latent log-volatilities; latent[,, r]
is the r
th draw
mu
Matrix of means of the log-volatilities; mu[, r]
is the r
th draw
phi
Matrix of AR(1) parameters for the log-volatilities; phi[, r]
is the r
th draw
sigma
Matrix of innovation variances for the log-volatilities; sigma[, r]
is the r
th draw
Ankargren, S., Unosson, M., & Yang, Y. (2020) A Flexible Mixed-Frequency Bayesian Vector Autoregression with a Steady-State Prior. Journal of Time Series Econometrics, 12(2), 10.1515/jtse-2018-0034. Ankargren, S., & Jon<U+00E9>us, P. (2020) Simulation Smoothing for Nowcasting with Large Mixed-Frequency VARs. Econometrics and Statistics, 10.1016/j.ecosta.2020.05.007. Ankargren, S., & Jon<U+00E9>us, P. (2019) Estimating Large Mixed-Frequency Bayesian VAR Models. arXiv:1912.02231, https://arxiv.org/abs/1912.02231. Kastner, G., & Huber, F. (2020) Sparse Bayesian Vector Autoregressions in Huge Dimensions. Journal of Forecasting, 39, 1142--1165. 10.1002/for.2680. Schorfheide, F., & Song, D. (2015) Real-Time Forecasting With a Mixed-Frequency VAR. Journal of Business & Economic Statistics, 33(3), 366--380. 10.1080/07350015.2014.954707
set_prior
, update_prior
, predict.mfbvar
, plot.mfbvar_minn
,
plot.mfbvar_ss
, varplot
, summary.mfbvar
# NOT RUN {
prior_obj <- set_prior(Y = mf_usa, n_lags = 4, n_reps = 20)
mod_minn <- estimate_mfbvar(prior_obj, prior = "minn")
# }
Run the code above in your browser using DataLab