VARsignR (version 0.1.3)

uhlig.reject: Uhlig's (2005) rejection method

Description

Identifies stuctural shocks using Uhlig's (2005) rejection method and estimates a Bayesian vector autoregression model with a flat Normal inverted-Wishart prior.

Usage

uhlig.reject(Y=NULL, nlags=4, draws=200, subdraws=200, nkeep=1000, KMIN=1, KMAX=4, constrained=NULL, constant=TRUE, steps=24)

Arguments

Y

A ts object containing the data series used for estimation; this should be of size $T x nvar$.

nlags

The number of lags to include of each variable. The default value is 4.

draws

An integer value for the number of Markov Chain Monte Carlo (MCMC) sampling replications. The default value is 200.

subdraws

An integer value for the number of sub-draws over the rejection routine. The default value is 200.

nkeep

An integer value for the number of desired MCMC draws that meet the imposed sign restrictions. The default value is 1000.

KMIN

An integer value for the first period of the impulse responses to which the sign restrictions apply. The default value is 1.

KMAX

An integer value for the last period of the impulse responses to which the sign restrictions apply. The default value is 4.

constrained

A list of sign restrictions of length $

constant

A logical statement on whether to include an intercept in the model. The default is 'TRUE'.

steps

An integer value for the horizon of the impulse response calculations. The default value is 24.

Value

A list of the posterior draws, which contains:

IRFS

A $draws x steps x nvar$ array of posterior impulse responses.

FEVDS

A $draws x steps x nvar$ array of posterior forecast error variance decompositions.

SHOCKS

A $draws x (T-nlags)$ array of posterior draws of the model's shocks.

Bdraws

A $draws x nvar$ array of posterior draws of the model coefficients.

Sdraws

A $draws x nvar x nvar$ array of posterior draws of the model's variance-covariance matrix.

Details

This function estimates a BVAR model using a flat Normal inverted-Wishart prior. Structural shocks are identified using Uhlig's (2005) rejection method. The size of the shock is one standard deviation. uhlig.reject is a partial identification method. Only one shock of interest is identified at a time by the sign restrictions provided. The MCMC routine stops when either enough draws have been found that satisfy the sign restrictions (specified in nkeep) or the maximum number of draws has been reached.

Sign restrictions are specified as follows: The first entry of constrained is the shock of interest corresponding to the column of the variable in $Y$. You MUST specify a sign restriction for the shock of interest on itself. Restrictions of the shock of interest on other variables are optional. The sign of the entry in the list corresponds to the direction of the sign, e.g. >0 or <0. variables="" that="" are="" not="" listed="" in="" constrained remain unconstrained. In the example below, the shock of interest is the 4th variable (FED funds rate) in $Y$. The sign indicates that the effect on itself is contrained to be positive. The responses of the 2nd, 3rd, and 5th variable in the model to a positive FED funds rate shock is restricted to be negative. The responses of the 1st and 6th variable are unconstrained.

References

Uhlig, H. (2005), "What Are the Effects of Monetary Policy on Output? Results from an Agnostic Identification Procedure", Journal of Monetary Economics, 52, 381-419.

See Also

rfbvar, uhlig.penalty, rwz.reject, fp.target, irfplot, fevdplot.

Examples

## Not run: 
# # Replication of Figure 6 of Uhlig (2005)
# 
# set.seed(12345)
# data(uhligdata)
# 
# # variable labels for plots
# vl <- c("GDP","GDP Deflator","Comm.Pr.Index","Fed Funds Rate",
#         "NB Reserves", "Total Reserves")
# 
# # sign restrictions
# # shock of interest enters first.
# # you MUST provide a restriction for the shock of interest
# # restriction variable 4 is >0
# # 2nd, 3rd, and 5th variable are <0.
# # 1st and 6th variable are unrestricted
# constr <- c(+4,-3,-2,-5)
# 
# # estimates the model
# model1 <- uhlig.reject(Y=uhligdata, nlags=12, draws=200, subdraws=200, nkeep=1000, KMIN=1,
#                         KMAX=6, constrained=constr, constant=FALSE, steps=60)
# 
# # get posterior draws
# irfs1 <- model1$IRFS
# 
# # plot impulse response functions
# irfplot(irfdraws=irfs1,type="median", labels=vl, save=FALSE, bands=c(0.16, 0.84),
#         grid=TRUE, bw=FALSE)
# 
# ## End(Not run)