Improve error messages for missing exogenous regressors in forecast's new_data about fabletools

[SomeoneSerge]

It seems, forecast.TSLM ain't even being run: error probably occurs somewhere in forecast.model which is defined I'm yet to learn where

[mitchelloharawild]

This is because you are using an exogenous regressor in your model (Const). When forecasting with exogenous regressors, it is not possible to make predictions without knowing future values of the regressors.

In this case, you would need to provide a tsibble to forecast() containing future time indices to be forecasted, along with the future values of Const.

I'm moving this to fablelite as an improvement to the error message.

For the example above:

library(fable)
library(tsibble)
library(tsibbledata)
library(lubridate)

# Adding a new measurable
aus_retail <-  aus_retail %>% mutate(Const=1)
future_retail <- new_data(aus_retail, n = 24) %>% mutate(Const=1)
aus_retail %>%
  filter(
    State %in% c("New South Wales", "Victoria"),
    Industry == "Department stores"
  ) %>% 
  model(
    lm = TSLM(Turnover~0+Const),
  ) %>%
  forecast(future_retail) %>% 
  autoplot(filter(aus_retail, year(Month) > 2010), level = NULL)

^{Created on 2019-03-21 by the reprex package (v0.2.1)}

[mitchelloharawild]

Error is now more informative.

library(tsibble)
library(tsibbledata)
library(lubridate)
library(fable)

aus_retail <-  aus_retail %>% mutate(Const=1)
aus_retail %>%
  filter(
    State %in% c("New South Wales", "Victoria"),
    Industry == "Department stores"
  ) %>% 
  model(
    lm = TSLM(Turnover~0+Const),
  ) %>%
  forecast %>% 
  autoplot(filter(aus_retail, year(Month) > 2010), level = NULL)
#> Error: object 'Const' not found
#> Unable to compute required variables from provided `new_data`.
#> Does your model require extra variables to produce forecasts?

^{Created on 2019-03-26 by the reprex package (v0.2.1)}