Implement function to grab the most recent data, sufficient for prediction

[dajmcdon]

For example, if we have max_lag = 14, then we would do something like:

x_latest <- x %>%
  filter(!is.na(case_rate), !is.na(death_rate)) %>%
  group_by(geo_value) %>%
  slice_tail(n = 15) # have lag 0,...,14, so need 15 for a complete case

The function should read the epi_recipe, perhaps after prep() or bake().

[ChloeYou]

Hi Daniel! @dajmcdon I was wondering if I can I get some clarifications on this?

Would this function be a separate function stored in the arx_forecaster.R file? Additionally, would this function go somewhere before line 44? or perhaps it should exist in another file?

Thank you in advance!

[ChloeYou]

new file, new function. get_target()/ get_test_data()
get recipe and get the data, and read the ahead (I think it should read the lag?)

example: look at

epipredict/musings/example-recipe.R

Line 39 in e8cfd8e

step_epi_ahead(death_rate, ahead = 7) %>%

b <- bake(prep(r))

[ChloeYou]

Logan suggested taking a look at the following that included this operation:

cmu-delphi/covidcast/R-packages/modeltools

create_train_and_predict_matrices() although it had issues when dealing with data with nonzero reporting latency

[ChloeYou]

TO-DO: Add unit tests for this new function before creating a PR