Handle missing values in NNETAR & VAR

[jonekeat]

Hi @mitchelloharawild,

Is it possible to handle missing values in NNETAR & VAR as in ARIMA? as I try to fit multiple models and compare them using accuracy() but I got error message like

which only work fine for ARIMA & TSLM

Here is the reproducible code:
xy <- tsibble::tsibble(date = tsibble::yearquarter(seq.Date(as.Date("1992-03-01"), by = "quarters", length.out = 111)), y = rnorm(111), x = c(rep(NA, 27), rnorm(111-27)))
xy %>% model(arima = ARIMA(y ~ x), tslm = TSLM(y ~ x), nnetar = NNETAR(y ~ x), var = VAR(y ~ x) ) %>% accuracy()

Here is my sessionInfo():

[mitchelloharawild]

Should be better now - thanks!