R/06GVAR_GC.R

In GVARX: Perform Global Vector Autoregression Estimation and Inference

.grangerGVAR <- function (data,p=2,FLag=3, type="const",lag.max=NULL, ic="AIC", weight.matrix=NULL, boot.no=0) {
ID <- NULL
  stopifnot(ncol(data)==4)
ic=ic
 p=p
 type=type

 idCol=which(colnames(data) == "ID")
 timeCol=which(colnames(data)=="Time")
 variates=colnames(data[,-c(idCol,timeCol)]) #names of column variables
 dat=data[,-timeCol] #Data with ID column only
 NAME=as.character(unique(dat[,"ID"]) )
 N=length(NAME) #number of countries

 timeID=as.character(subset(data, ID==NAME[1])[,timeCol])
 Year=unique(as.character(lubridate::year(timeID)))

output1=output2=output3=outRSD=EXO=list()
Y1gcY2=Y2gcY1=NULL
Y1gcY2.GVAR=Y2gcY1.GVAR=NULL

Ft=GVAR_Ft(data,weight.matrix)

for (i in 1:N)  {
 exo=embed(as.matrix(Ft[[i]]),FLag)   # Foreign variables
 ytmp=as.matrix(subset(data,ID==NAME[i])[,-c(1:2)])
 varnames=paste(NAME[i],paste(".",variates,sep=""),sep="")
colnames(ytmp)=varnames

yL=as.matrix(ytmp[-c(1:(FLag-1)),])
print(nrow(yL))
print(nrow(exo))
exoNAMES=NULL
for (j in 1:FLag) {
exoNAMES=c(exoNAMES,paste("F.",variates,".Lag",j-1,sep=""))
}

colnames(exo)=c(exoNAMES)

##==Compute Granger Causality based upon ordinary VAR
lag.max=lag.max
VAR.GC=vars::VAR(y=yL, p = p, type = type, lag.max = lag.max, ic=ic)
if (boot.no==0) {
GCY1=vars::causality(VAR.GC,cause=varnames[1],vcov.=vcovHC(VAR.GC))$Granger
GCY2=vars::causality(VAR.GC,cause=varnames[2],vcov.=vcovHC(VAR.GC))$Granger } else {
GCY1=vars::causality(VAR.GC,cause=varnames[1],boot = TRUE, boot.runs = boot.no)$Granger
GCY2=vars::causality(VAR.GC,cause=varnames[2],boot = TRUE, boot.runs = boot.no)$Granger}

Y1gcY2_tmp=cbind(as.numeric(GCY1$statistic[1]),as.numeric(GCY1$p.value))
Y2gcY1_tmp=cbind(as.numeric(GCY2$statistic[1]),as.numeric(GCY2$p.value))
Y1gcY2=rbind(Y1gcY2,Y1gcY2_tmp)
Y2gcY1=rbind(Y2gcY1,Y2gcY1_tmp)

##==Compute Granger Causality based upon GVARX
GVAR.GC=vars::VAR(y=yL, p = p, type = type, exogen = exo, lag.max = lag.max, ic=ic)

if (boot.no==0) {GCY1.GVAR=vars::causality(GVAR.GC,cause=varnames[1],vcov.=vcovHC(GVAR.GC))$Granger
GCY2.GVAR=vars::causality(GVAR.GC,cause=varnames[2],vcov.=vcovHC(GVAR.GC))$Granger
} else {
GCY1.GVAR=vars::causality(GVAR.GC,cause=varnames[1], boot = TRUE, boot.runs = boot.no)$Granger
GCY2.GVAR=vars::causality(GVAR.GC,cause=varnames[2], boot = TRUE, boot.runs = boot.no)$Granger
}

Y1gcY2.GVAR_tmp=cbind(as.numeric(GCY1.GVAR$statistic[1]),as.numeric(GCY1.GVAR$p.value))
Y2gcY1.GVAR_tmp=cbind(as.numeric(GCY2.GVAR$statistic[1]),as.numeric(GCY2.GVAR$p.value))
Y1gcY2.GVAR=rbind(Y1gcY2.GVAR,Y1gcY2.GVAR_tmp)
Y2gcY1.GVAR=rbind(Y2gcY1.GVAR,Y2gcY1.GVAR_tmp)

} # end of 1 st jj countryloop

options(digits=4)
## Reporting Granger Causality output based on VAR
rownames(Y1gcY2)=NAME
null1=paste(variates[1],"does not Granger cause",variates[2])
colnames(Y1gcY2)=c(null1,"pvalue, VAR")

rownames(Y2gcY1)=NAME
null2=paste(variates[2],"does not Granger cause",variates[1])
colnames(Y2gcY1)=c(null2,"pvalue, VAR")


## Reporting Granger Causality output based on GVAR
rownames(Y1gcY2.GVAR)=NAME
null1=paste(variates[1],"does not Granger cause",variates[2])
colnames(Y1gcY2.GVAR)=c(null1,"pvalue, GVAR")

rownames(Y2gcY1.GVAR)=NAME
null2=paste(variates[2],"does not Granger cause",variates[1])
colnames(Y2gcY1.GVAR)=c(null2,"pvalue, GVAR")

#print(Y1gcY2);print(Y2gcY1)
#print(Y1gcY2.GVAR);print(Y2gcY1.GVAR)

results <- list(y1GCy2.var=Y1gcY2,y2GCy1.var=Y2gcY1,y1GCy2.gvar=Y1gcY2.GVAR,y2GCy1.gvar=Y2gcY1.GVAR)

return(results)

} ## end of function GC()