GVARX (version 1.4)

GVAR_GF: Compute the structural coefficients matrices G0, G1, G2, and F1, F2

Description

Compute the structural coefficients matrices G0, G1, G2, and F1, F2

Usage

GVAR_GF(data,p, type="const",ic="AIC",weight.matrix)

Value

G0

Matrix G0 of Eq.(2.6) in Filippo and Pesaran(2013, P.17)

G1

Matrix G1 of Eq.(2.6) in Filippo and Pesaran(2013, P.17)

G2

Matrix G2 of Eq.(2.6) in Filippo and Pesaran(2013, P.17)

F1

Matrix F1 of Eq.(2.6) in Filippo and Pesaran(2013, P.17)

F2

Matrix F2 of Eq.(2.6) in Filippo and Pesaran(2013, P.17)

lagmatrix

Country-secific optimal lag number, which must be the same.

RESID

original residuals=u in Filippo and Pesaran (2013, P.17)

newRESID

New residuals=epsilon in Filippo and Pesaran (2013, P.17)

fitted

In-sample fitted values, or conditional mean

data

data used

Arguments

data

Dataframe is a strictly balanced panel data format,the first column is cross-section ID,and the second column is Time. For the sake of identification, both columns must be named by, respectively, id and Time.

p

The number of lag for Xt matrix. The number of lag for foreign variables in country-specific VAR FLag is set to be p+1. Current version restricts p <= 2 for simplicity, which aims at avoiding too many paramaters in low-frequency data of many variables and many countries. It will be relaxed soon.

type

Model specificaiton for VAR. As in package vars, we have four selection: "none","const","trend", "both".

ic

Information criteria for optimal lag.As in package vars, we have four selection: "AIC", "HQ", "SC", "FPE".

weight.matrix

Bilateral trade weight matrix for computing foreign variables. If the computation of foreign variables are weighted by one weighting matrix, weight.matrix must be a "data.frame". If the computation of foreign variables are weighted on a year-to-year basis, then weight.matrix must be a "list", with the same length as the weighting frequency.

Author

Ho Tsung-wu <[e-mail]>, College of Management, National Taiwan Normal University.

Details

This function generates several structural coefficient matrices of Eq.(2.6) in Filippo and Pesaran(2013, P.17), which are required to compute IRF and multistep forecasts. Besides, it also re-calculates the transformed residuals. In this version, we do not include the impulse responses function(IRF), because the IRF can be computed by these matrices and residuals easily. We will not update it until the next version.

References

Mauro Filippo di and Pesaran H. M. (2013) The GVAR Handbook-- Structure and Applications of a Macro Model of the Global Economy for Policy. Oxford University Press.

Examples

data("PriceVol")
data("tradeweightx")
data("tradeweight1")
p=2
type="const"
ic="SC"

Result=GVAR_GF(data=PriceVol,p,type,ic, weight.matrix=tradeweight1)
Result$G0
Result$G1
Result$G2
Result$F1
Result$F2
Result$lagmatrix
Result$RESID
Result$newRESID
Result$fitted
Result$data
#May use forecast::accuracy(Result$fitted[,1], Result$data[,1]) for performance.