Many parameters have placeholder documentation

[brookslogan]

E.g., in ?DelphiRF::DelphiRF:

train_models-template: DESCRIPTION GOES HERE

make_predictions-template: DESCRIPTION GOES HERE

model_save_dir-template: DESCRIPTION GOES HERE

indicator-template: DESCRIPTION GOES HERE

lp_solver-template: DESCRIPTION GOES HERE

signal-template: DESCRIPTION GOES HERE

geo_level-template: DESCRIPTION GOES HERE

geo-template: DESCRIPTION GOES HERE

signal_suffix-template: DESCRIPTION GOES HERE

lambda-template: DESCRIPTION GOES HERE

gamma-template: DESCRIPTION GOES HERE

value_type-template: DESCRIPTION GOES HERE

training_end_date-template: DESCRIPTION GOES HERE

training_days-template: DESCRIPTION GOES HERE

It's possible that this is the result of something going wrong migrating to roxygen @templates; it looks like many/all of the template files here have placeholder text, while the parameters that do not involve templates have working documentation.

[brookslogan]

Some other entries in the README and/or @param documentation that could be clarified:

refd_col: Character specifying the reference column for merging operations.

should probably be "reference date column"

ref_lag: Integer specifying the reference lag value.

I'm guessing this is specifying the target lag that will be predicted?

df: A data frame containing historical data with 'target_date' and 'report_date' columns.

It would help to clarify that this should contain an additional column (or columns?) containing time series data as it was reported on each report date. Some nowcasting packages are built with nondecreasing count revisions in mind, and they instead have the number of events newly reported at each reference date x report date instead. Or they take line lists as input.

testing_start_date: A Date object indicating the start date for testing (e.g., as.Date("2022-06-01")).

I don't know what this means. Is it a reference date or a report date? How is it used? --- In DelphiRF, it seems to be used as a target reference date, used for selecting report dates for the test set as well as determining a training window: (testing_start_date - training_days, testing_start_date]. In revision_forecast, it seems to be used as a report date instead, used for a one-sided window (ignoring training_days). The NA target_dates introduced by data_preprocessing when we don't have ref_lag-mature data available makes placing testing_start_date at the nowcast date (and for weekly, pre-aligning reference dates and report dates to lie on the same wday, in a way such that lag 0 "makes sense") are already essentially doing one side of the DelphiRF filter for a testing_start_date of the nowcast date (dplyr::filter turns NA comparisons into exclusions), so setting testing_start_date to the (aligned) nowcast date feels natural; testing_start_date = aligned nowcast date is also very natural for the revision_forecast report date filter. [This is somehow wrong, making the testing start date later than the nowcast date seems to potentially make more training data available, though it may mean that 0 prediction rows are produced, not sure.]

test_lag_groups: A list specifying the lag groups to be tested (default: TEST_LAG_GROUPS).

Should be referencing TEST_LAG_GROUPS_DAILY and TEST_LAG_GROUPS_WEEKLY.

smoothed: Logical indicating whether smoothing should be applied.

and

smoothed: A logical value (TRUE/FALSE) indicating whether the value column is alrealdy smoothed (default: FALSE)

are saying two different things. Only the latter makes sense for a parameter called smoothed.

smoothed_target: A logical value indicating whether the target variable should be smoothed.

I don't know if this means I should have already smoothed it / data_preprocessing should have already smoothed it, or if this is controlling whether DelphiRF/revision_forecast will smooth it.

params_list: doesn't say what parameters it's controlling in ?DelphiRF or ?revision_forecast.

signal: A string specifying the signal name associated with the indicator (default: "").

I don't know what this means vs. the indicator parameter. (Is this like "source" and "signal" in the Epidata API?)

lag_pad: A numeric value indicating the padding for lag variables (default: 1).

I don't know what this means.