Bugs in modeltools::quantgen_forecaster

[kjytay]

I found some bugs when trying to run modeltools::quantgen_forecaster(). First issue I'm raising here, so not sure if better to put them together or to separate.

1. df must be a list, if not line 152 (call to covidcast::aggregate_signals()) will fail. This is a problem if df is the returned object from covidcast::covidcast_signals() where only one signal was retrieved. In that case, the returned object from covidcast_signals() is not a list.

There are a few ways to fix this. Checking if df is a list and if not, make it so is probably the best solution.

Reproducible example:

df <- covidcast::covidcast_signal(data_source = "usa-facts", 
                                  signal = "confirmed_7dav_incidence_prop",
                                  start_day = "2020-08-01", 
                                  end_day = "2020-08-15")

modeltools::quantgen_forecaster(
  df,
  forecast_date = "2020-08-16", 
  signals = tibble::tibble(
    data_source = c("usa-facts"),
    signal = c("confirmed_7dav_incidence_prop")
  ), 
  incidence_period = "day",
  ahead = 0,
  n = 15,
  verbose = TRUE, 
  lambda = 0)

#Error in covidcast::aggregate_signals(df, dt = dt, format = "wide") : 
#  If `x` is a `covidcast_signal` data frame, then `dt` must be a vector.

2. The code below results in an error:

# Same df as in Bug 1
df <- covidcast::covidcast_signal(data_source = "usa-facts", 
                                  signal = "confirmed_7dav_incidence_prop",
                                  start_day = "2020-08-01", 
                                  end_day = "2020-08-15")

# Same call as in Bug 1 except for `df` argument
modeltools::quantgen_forecaster(
  list(df),
  forecast_date = "2020-08-16", 
  signals = tibble::tibble(
    data_source = c("usa-facts"),
    signal = c("confirmed_7dav_incidence_prop")
  ), 
  incidence_period = "day",
  ahead = 0,
  n = 15,
  verbose = TRUE, 
  lambda = 0)

# ahead = 0Error in apply(x, 2, mean) : dim(X) must have a positive length

This error goes away when I specify lags = c(0, 3), or even lags = 1. But when I include these lags options, the function call never terminates and I have to restart RStudio.

3. The geo_type argument is not used anywhere in the function and should probably be removed.