I use the updated greybox package in R to forecast the consecutive 2 values (horizon=2) with a moving average scheme (See the first line of code below), where the window size is equal to 3.
For example, the overall goal is to take the average of (1+2+3)/3 = "2" as the forecasted value in horizon 1 (h=1) and then make use of the predicted value in h=1 for h=2, where (2+3+"2")=2,3334. The following forecast window will make use of the window (2+3+4), where 4 is the actual value to predict the next h1 and h2, which equals 3 and 3,3334 respectively.
Yet, the prediction result I want "ValuesMA[[3]]" only emits one row, i.e. values for the first horizon. But it should be equal to the predifined horizon, which is two.
I have a code for an AR1 process which works perfectly (Second line of code). At the end I add an MAE test statistic to evaluate the model.
Can anyone help?
Thank you!
This is the underlying code I use:
#data
z <- c(1,2,3,4,5,6,7)
ourCall <- "mean(x=data,n.ahead=h)"
ourValue <- c("pred")
# Return a list for an forecasting horizon h
ValuesMA <- ro(z, h=2, origins=3, call=ourCall, ci=TRUE, co=TRUE)
ValuesMA[[3]]
**Which yields:**
origin3 origin4 origin5
[1,] 2 3 4
**But I want:**
origin3 origin4 origin5
[1,] 2 3 4
[2,] 2,3334 3,3334 4,3334
#data
z <- c(1,2,3,4,5,6,7)
# ci defines constant in-sample window size, co defines whether the holdout sample window size should be constant
ourCall <- "predict(arima(x=data,order=c(1,0,0)),n.ahead=h)"
# Ask for predicted values and the standard error
ourValue <- c("pred","se")
# Return a list for an forecasting horizon h with a rolling holdout size equal to origin
ValuesAR1 <- ro(z, h=2, origins=3, call=ourCall, value=ourValue, ci=TRUE, co=TRUE)
# calculate MAE
MAE_AR1 <- apply(abs(ValuesAR1$holdout - ValuesAR1$pred),1,mean,na.rm=TRUE) / mean(ValuesAR1$actuals)
ValuesAR1[[3]]
**Which yields:**
> ValuesAR1[[3]]
origin3 origin4 origin5
h1 2 3 4
h2 2 3 4
For further reading see: https://cran.r-project.org/web/packages/greybox/vignettes/ro.html
