use_ensemble_criteria: Load models according to inclusion criteria before ensembling

In epiforecasts/forecasthubutils2: What the Package Does (One Line, Title Case)

Load models according to inclusion criteria before ensembling

Description

Load models according to inclusion criteria before ensembling

Usage

use_ensemble_criteria(
  forecasts,
  exclude_models = NULL,
  exclude_designated_other = TRUE,
  return_criteria = TRUE,
  eval_dir = here::here("evaluation", "weekly-summary"),
  rel_wis_cutoff = Inf,
  config_file = here::here("project-config.json")
)

Arguments

forecasts	data.frame containing all the forecasts to be summarised as an ensemble.
exclude_models	optional character vector to exclude over all dates, or data.frame with cols model and forecast_date, to exclude for specific dates
exclude_designated_other	logical: whether to exclude models designated as "other" in their metadata file (default TRUE)
return_criteria	logical : whether to return a model/inclusion criteria grid as well as the ensemble forecast (default TRUE)
eval_dir	character: the path in which to look for evaluation csv files
rel_wis_cutoff	numeric: any model with relative WIS greater than this value will be excluded
config_file	Path to the config file

Details

Steps: Currently, models included based on having:

1. All quantiles

2. 4 horizons

3. Not manually specified for exclusion

4. Not the hub ensemble

Value

• if return_criteria = TRUE, a list with the following elements

  • "ensemble" : tibble : a single ensemble forecast

  • "criteria": tibble : all candidate models against criteria for inclusion in ensemble (all locations and horizons)

  • "forecast_date" : date : latest date

• if return_criteria = FALSE, a tibble of a single ensemble forecast

epiforecasts/forecasthubutils2 documentation built on July 1, 2023, 11:50 a.m.