Bug: get_test_data uses na.omit() on all data table columns, rather than just predictors/outcomes

[dshemetov]

The line in question is here.

I realize we have consensus that get_test_data() will be re-written, but

• it's been that way for over a year now, so a temp fix might still be worth it here
• if a temp fix is not worth it, figured I'd report this so we don't re-create this bug in the new implementation

(Sorry this reprex isn't very minified)

library(epipredict)
library(epiprocess)
library(epidatr)
library(data.table)
library(dplyr)
library(tidyr)
library(ggplot2)
library(magrittr)
library(purrr)

#Fetch data from Delphi Epidata API 
nhsn <- pub_covidcast(
  source = "nhsn",
  signals = "confirmed_admissions_covid_ew_prelim",
  geo_type = "state",
  time_type = "week",
  geo_values = "*",
  time_values = epirange(202001, 202601),
  issues = epirange(202001, 202601)
) |>
  # The version date column is called `issue` in the Epidata API. Rename it.
  rename(version = issue, confirmed_covid_adm = value) |>
  as_epi_archive(compactify = TRUE)

nhsn_original <- nhsn
nhsn <- nhsn_original
#Plot of monthly snapshots of the data
geo_choose <- "ca"

forecast_dates <- as.Date("2025-03-09")

covid_adm_data <- bind_rows(
  # Snapshotted data for the version-faithful forecasts
  map(
    forecast_dates,
    ~ nhsn |>
      epix_as_of(.x) |>
      mutate(version = .x)
  ) |>
    bind_rows() |>
    mutate(version_faithful = "Version faithful"),
  # Latest data for the version-un-faithful forecasts
  nhsn |>
    epix_as_of(nhsn$versions_end) |>
    mutate(version_faithful = "Version un-faithful")
)

# Enter debug mode here
debug(epipredict::get_test_data)
archive_cases_dv_subset_faux %>%
  epix_as_of(as.Date("2025-03-09")) %>%
  arx_forecaster("confirmed_covid_adm", "confirmed_covid_adm", args_list = arx_args_list(
          ahead = 7,
          lags = c(0, 7, 14, 21),
          adjust_latency = "extend_ahead"
        ))