out-of-sample VAR(1) rolling window, mean absolute error

Question

Please kindly help, any advise counts thanks!

I am trying to calculate the MAE (Mean Absolute Error) for VAR(1) rolling window but cannot resolve the error. Code:

RRECG <- data.frame(na.omit(cbind(R1,R2,E1,C1,G1)))
set.seed(123)
y<-Df$R1
S=171;h=113;
error1.h<-c()
for (i in S:(length(y)-h))
{
  VARO.sub<- VAR(na.omit(RRECG[(1:i),],p=1,type="const"))
  predict.h1<-predict(VARO.sub,n.ahead=h)
  error1.h<-c(error1.h,y[i+h]-predict.h1)
}
summary(abs(error1.h))

Error message is:

Error in y[i + h] - predict.h1 : non-numeric argument to binary operator

Tried:

Also tried:
``` predict.h1<-as.numeric(predict(VARO.sub,n.ahead=h))
But not even close to resolve the issue. Please kindly help, thank you!