How to use cvts on an hybridModel with XREG? - r

Question

I am trying to do a cross validation using the cvts function from the forecastHybrid package using an "an" model (ARIMA + NNETAR) with external regressors.
I have two variables with 100 observations : Y and X
Note that:
length(Y) == length(X)
TRUE
I did this:
crossv =cvts(Y,
FUN=hybridModel, models="an", a.args=list(xreg=X,n.args=list(xreg=X),
rolling = TRUE, windowSize = 84, maxHorizon = 1, horizonAverage = FALSE)
and got this error
Error in { :
task 1 failed - "variable lengths differ (found for 'xregg')"
if I try to pass it as a function.
CUSTOM=function(x){hybridModel(x, models="an", a.args=list(xreg=X),n.args=list(xreg=X))}
crossv2 = cvts(Y,
FUN=CUSTOM,
rolling = TRUE,
windowSize = 84,
maxHorizon = 1,
horizonAverage = FALSE)
I get:
Error in { : task 1 failed - "object 'X' not found"
I can, of course do cross validation separately for nnetar and arima and then averaging the cross validation forecasting but any ideas why it’s not working via cvts + hybrid model ?
Thanks a lot.

Answer 1

I finally found the answer. We should add an other xreg argument in the FUN function. Then, a second xreg argument (from the cvts function) should be written outside the function. The value of the second xreg are then passed to the first xreg.
y = ts(rnorm(100), start = c(1999,1), frequency = 12)
x = ts(rnorm(100), start = c(1999,1), frequency = 12)
cv = cvts(y, FUN = function(z, xreg = xreg),
forecastHybrid::hybridModel(z,models = "an")},
xreg=as.matrix(x),rolling = TRUE, windowSize = 84,
maxHorizon = 1, horizonAverage = FALSE)
It worked!
accuracy(cv)
ME RMSE MAE
Forecast Horizon 1 -0.2173456 0.9328878 0.7368945