calc_model_eligibility_for_ensemble: Generate a data frame with a row for each model and an...

In reichlab/covidEnsembles: What the Package Does (One Line, Title Case)

Generate a data frame with a row for each model and an indicator of whether that model is eligible for inclusion in an ensemble, with explanation if not

Description

Generate a data frame with a row for each model and an indicator of whether that model is eligible for inclusion in an ensemble, with explanation if not

Usage

calc_model_eligibility_for_ensemble(
  qfm,
  observed_by_location_target_end_date,
  missingness_by_target = FALSE,
  do_q10_check = TRUE,
  do_nondecreasing_quantile_check = TRUE,
  do_baseline_check = FALSE,
  do_sd_check = "exclude_none",
  sd_check_table_path = NULL,
  sd_check_plot_path = NULL,
  window_size = 0,
  decrease_tol = 1,
  baseline_tol = 1.2
)

Arguments

qfm	matrix of model forecasts of class QuantileForecastMatrix
observed_by_location_target_end_date	data frame of observed values
missingness_by_target	logical; if TRUE, check condition that forecasts are not missing for each combination of model, week, and target; otherwise,
do_q10_check	logical; if TRUE, check condition that quantile at level 0.1 is at least as large as the most recent observed value
do_nondecreasing_quantile_check	logical; if TRUE, check condition that quantiles for consecutive targets (1 wk ahead, 2 wk ahead, etc) are non-decreasing for each combination of location, forecast_week_end_date, model, and quantile probability level
do_baseline_check	logical; if TRUE, check condition that WIS for model is within specified tolerance of WIS for baseline
do_sd_check	character: "exclude_none" to keep all forecasts, "exclude_below" to exclude low forecasts, "exclude_above" to exclude high forecasts, or "exclude_both" to exclude both low and high forecasts. exclusions are based on a check of whether the mean of the next (7) predicted medians is not more than (4) sample standard deviations abov or below the mean of the last (14) observations
window_size	non-negative integer number of historic weeks that are examined for forecast missingness; 0 is appropriate for equal weight ensembles where no historical data is required. If two past weeks of forecast data are required to estimate ensemble parameters, window_size should be 2
decrease_tol	numeric; decreases of up to specified tolerance are allowed
baseline_tol	numeric; for baseline check, model's mean WIS for forecasts within the window size must be at most baseline_tol times the mean wis for the baseline model on the corresponding forecasts

Value

data frame with two columns:

• one with name given by model_id_name recording model id for each model

• a second called 'eligibility' that is either the string 'eligible' or a brief description of why the model can't be included

reichlab/covidEnsembles documentation built on Jan. 31, 2024, 7:21 p.m.