load_forecasts: Load all available forecasts submitted around forecast dates...

In reichlab/covidHubUtils: Utility functions for the COVID-19 forecast hub

Load all available forecasts submitted around forecast dates from Zoltar, local Zoltar module or local hub repo.

Description

• If date_window_size is 0, this function returns all available forecasts submitted on every day in dates.

• Ifdate_window_size is not 0, this function will look for the latest forecasts that are submitted within window size for each day in dates.

• If source is local_zoltar, a valid sqlite3 object is required. Please follow the instruction in load_forecasts_local_zoltar() to set up required environment.

Usage

load_forecasts(
  models = NULL,
  dates = NULL,
  date_window_size = 0,
  locations = NULL,
  types = NULL,
  targets = NULL,
  source = "zoltar",
  hub_repo_path,
  local_zoltpy_path,
  zoltar_sqlite_file,
  data_processed_subpath = "data-processed/",
  as_of = NULL,
  hub = c("US", "ECDC", "FluSight"),
  verbose = TRUE
)

Arguments

models	Character vector of model abbreviations. Default all models that submitted forecasts meeting the other criteria.
dates	The forecast date of forecasts to retrieve. A vector of one or more Date objects or character strings in format “YYYY-MM-DD” Default to all valid forecast dates.
date_window_size	The number of days across each date in dates parameter to look for the most recent forecasts. Default to 0, which means to only look at the dates parameter only.
locations	a vector of strings of fips code or CBSA codes or location names, such as "Hampshire County, MA", "Alabama", "United Kingdom". A US county location names must include state abbreviation. Default to NULL which would include all locations with available forecasts.
types	Character vector specifying type of forecasts to load: "quantile" and/or "point". Default to all valid forecast types.
targets	character vector of targets to retrieve, for example c('1 wk ahead cum death', '2 wk ahead cum death'). Default to NULL which stands for all valid targets.
source	string specifying where forecasts will be loaded from: one of "local_hub_repo", "zoltar" and "local_zoltar". Default to "zoltar".
hub_repo_path	path to local clone of the forecast hub repository
local_zoltpy_path	path to local clone of zolpy repository. Only needed when source is ⁠"local_zoltar⁠.
zoltar_sqlite_file	path to local sqlite file, either a relative path w.r.t. local_zoltpy_path or an absolute path. Only needed when source is ⁠"local_zoltar⁠.
data_processed_subpath	folder within the hub_repo_path that contains forecast submission files. Default to "data-processed/", which is appropriate for the covid19-forecast-hub repository.
as_of	character for date time to load forecasts submitted as of this time from Zoltar. Ignored if source = "local_hub_repo". It could use the format of one of the three examples: ⁠"2021-01-01", "2020-01-01 01:01:01" and "2020-01-01 01:01:01 UTC"⁠. If you would like to set a timezone, it has to be UTC now. If not, helper function will append the default timezone to your input based on hub parameter. Default to NULL to load the latest version.
hub	character vector, where the first element indicates the hub from which to load forecasts. Possible options are "US", "ECDC" and "FluSight".
verbose	logical to print out diagnostic messages. Default is TRUE.

Value

data.frame with columns model, forecast_date, location, horizon, temporal_resolution, target_variable, target_end_date, type, quantile, value, location_name, population, geo_type, geo_value, abbreviation

Examples

# Load forecasts from US forecast hub
# This call only loads the latest forecast submitted on "2021-07-26" in
# a 12-day window w.r.t "2021-7-30".
load_forecasts(
  models = "COVIDhub-ensemble",
  dates = "2021-07-30",
  date_window_size = 11,
  locations = "US",
  types = c("point", "quantile"),
  targets = paste(1:4, "wk ahead inc case"),
  source = "zoltar",
  verbose = FALSE,
  as_of = NULL
)

# Load forecasts from ECDC forecast hub
# This function call loads the latest forecasts in each 2-day window
# w.r.t "2021-03-08" and "2021-07-27".
load_forecasts(
  models = "ILM-EKF",
  hub = c("ECDC", "US"),
  dates = c("2021-03-08", "2021-07-27"),
  date_window_size = 1,
  locations = "GB",
  targets = paste(1:4, "wk ahead inc death"),
  source = "zoltar"
)

reichlab/covidHubUtils documentation built on Feb. 6, 2024, 1:42 p.m.