Skip to contents

Forecast ILI

Source: R/forecast.R
forecast_ili.Rd

This function forecasts ILI up to a specified future horizon. The models used can be parameterized with a "models" argument (for more details see ts_fit_forecast). By default, the function will use an ARIMA approach to model all locations in the input historical ILI data and then use the fitted models to forecast out to each of the horizons.

Usage

forecast_ili(
  ilidat,
  horizon = 4L,
  trim_date = NULL,
  models = list(arima = "PDQ(0,0,0)+pdq(1:2,0:2,0)")
)

Arguments

ilidat

Data returned from get_cdc_ili

horizon

Optional horizon periods through which the forecasts should be generated; default is 4

trim_date

Earliest start date you want to use for ILI data; default NULL doesn't trim

models

The list of model parameters passed to ts_fit_forecast; defaults to list(arima="PDQ(0,0,0)+pdq(1:2,0:2,0)"

Value

A named list containing:

  • ilidat: The data sent into the function filtered to the location and the trim_date. Select columns returned.

  • ilidat_tsibble: The tsibble class object returned by running make_tsibble on the data above.

  • ili_fit: The fit from fabletools::model.

  • ili_forecast: The forecast from fabletools::forecast at the specified horizon.

  • ili_future: The horizon-number of weeks of ILI data forecasted into the future.

  • ili_bound: The data in 1 bound to the data in 5.

  • arima_params: A tibble with ARIMA model parameters for each location (if type="arima").

  • locstats: A tibble with missing data information on all locations.

  • removed: A tibble with locations removed because of high missing ILI data.

Examples

if (FALSE) { # \dontrun{
# Retrieve ILI data
ilidat <- get_cdc_ili(region = c("national", "state", "hhs"),
                      years = 2010:lubridate::year(lubridate::today()))

# Using data only from march 2020 forward, for US only
ilidat_us <- ilidat %>% dplyr::filter(location=="US")
# Replace most recent week with nowcast data, and nowcast last week
ilidat_us <- ilidat_us %>% replace_ili_nowcast(weeks_to_replace=1)
ilifor_us <- forecast_ili(ilidat_us, horizon=4L, trim_date="2020-03-01")
# Take a look at objects that come out ILI forecasting procedure
ilifor_us$ili_fit
ilifor_us$arima_params
ilifor_us$ili_forecast
head(ilifor_us$ili_bound)
tail(ilifor_us$ili_bound, 10)
} # }