dfev {MSBVAR}R Documentation

Decompositions of Forecast Error Variance (DFEV) for VAR models

Description

Computes the M dimensional decomposition of forecast error variance for a VAR model. User can specify the decomposition of the contemporaneous innovations.

Usage

dfev(var.obj, A0 = t(chol(var.obj$mean.S)), k)

Arguments

var.obj VAR object created from fitting a VAR model using either szbvar() or mlevar()
A0 Decomposition of the contemporaneous error covariance matrix. Default is to use the Cholesky decomposition of the residual error covariance matrix.
k Number of periods over which to compute the deccomposition.

Details

The decomposition of the forecast error variance (DFEV) provides a measure of the relationship among forecast errors or impact of shocks to a VAR system. It is computed by finding the moving average representation (MAR) of the VAR model and then tracing out the path of innovations through the system. For each of the M innovations in a VAR, the amount of the variance in these forecast errors or innovations is computed and returned in a table. The table can be accessed via the print.dfev and summary.dfev functions.

Value

Returns a list with

errors M x M x K of the percentage of the innovations in variable i explained by the other M variables.
std.err M x k dimension matrix of the forecast standard errors.
names Variable names

Note

The interpretation of the DFEV depends on the decomposition of the contemporaneous residuals. In the default case of a Cholesky decomposition, this means that the ordering of the variables in the decomposition determines the effect of each innovation on the subsequent DFEVs. For high correlated series, this will mean that the DFEV is not very robust to the ordering.

Author(s)

Patrick T. Brandt

References

See Also

See also print.dfev and summary.dfev for a nicely formatted tables and an output example

Examples

    data(IsraelPalestineConflict)
    varnames <- colnames(IsraelPalestineConflict)
    fitted.BVAR <- szbvar(IsraelPalestineConflict, p=6, z=NULL,
                           lambda0=0.6, lambda1=0.1,
                           lambda3=2, lambda4=0.25, lambda5=0, mu5=0,
                           mu6=0, nu=3, qm=4, prior=0,
                           posterior.fit=FALSE)

    A0 <- t(chol(fitted.BVAR$mean.S))
    dat.dfev <- dfev(fitted.BVAR, A0, 24)

    print.dfev(dat.dfev)
    summary.dfev(dat.dfev)

[Package MSBVAR version 0.1 Index]