This function serves as a plotting mechanism for prepped forecast submission data. The plots show the historical trajectory of the truth data supplied along with the forecasted point estimates and (optionally) the prediction interval. All plots are faceted by location.
Note that the ".data" and "submission" arguments to this function expect incoming data prepared in a certain format. See the argument documentation and "Details" for more information.
Usage
plot_forecast(
.data,
submission,
location = "US",
pi = 0.95,
.model = NULL,
.outcome = "flu.admits",
format = "legacy"
)Arguments
- .data
A data frame with historical truth data for all locations and outcomes in submission targets
- submission
Formatted submission (e.g., a
tibblecontaining forecasts prepped with format_for_submission)- location
Vector specifying locations to filter to;
'US'by default.- pi
Width of prediction interval to plot; default is
0.95for 95% PI; if set toNULLthe PI will not be plotted- .model
Name of the model used to generate forecasts; default is
NULLand the name of the model will be assumed to be stored in a column called "model" in formatted submission file- .outcome
The name of the outcome variable you're plotting in the historical data; defaults to
"flu.admits"- format
The submission format to be used; must be one of
"hubverse"or"legacy"; default is"legacy"
Details
To plot the forecasted output alongside the observed historical data, both the ".data" and "submission" data must be prepared at the same geographic and temporal resolutions. The data frame passed to ".data" must include the column specified in the ".outcome" argument as well as the following columns:
location: FIPS location code
week_end: Date of the last day (Saturday) in the given epidemiological week
If format is "legacy" the "submission" data should be a probabilistic forecast prepared as a tibble with at minimum the following columns:
forecast_date: Date of forecast
target: Horizon and name of forecasted target
target_end_date: Last date of the forecasted target (e.g., Saturday of the given epidemiological week)
location: FIPS code for location
type: One of either "point" or "quantile" for the forecasted value
quantile: The quantile for the forecasted value;
NAif "type" is"point"value: The forecasted value
If format is "hubverse" the "submission" data should be a probabilistic forecast prepared as a tibble with at minimum the following columns:
reference_date: Date of reference for forecast submission
horizon: Horizon for the given forecast
target: Name of forecasted target
target_end_date: Last date of the forecasted target (e.g., Saturday of the given epidemiological week)
location: Name or geographic identifier (e.g., FIPS code) for location for the given forecast
output_type: Type of forecasted value (e.g., "quantile")
output_type_id: The quantile for the forecasted value if output_type is "quantile"
value: The forecasted value
The "submission" data may optionally include a column with the name of the model used, such that multiple models can be visualized in the same plot.
Examples
if (FALSE) { # \dontrun{
# Get some data
h_raw <- get_hdgov_hosp(limitcols=TRUE)
# Prep all the data
prepped_hosp_all <- prep_hdgov_hosp(h_raw)
# What are the last four weeks of recorded data?
last4 <-
prepped_hosp_all %>%
dplyr::distinct(week_start) %>%
dplyr::arrange(week_start) %>%
tail(4)
# Remove those
prepped_hosp <-
prepped_hosp_all %>%
dplyr::anti_join(last4, by="week_start")
# Make a tsibble
prepped_hosp_tsibble <- make_tsibble(prepped_hosp,
epiyear = epiyear,
epiweek=epiweek,
key=location)
# Limit to just one state and US
prepped_hosp_tsibble <-
prepped_hosp_tsibble %>%
dplyr::filter(location %in% c("US", "51"))
# Fit models and forecasts
hosp_fitfor <- ts_fit_forecast(prepped_hosp_tsibble,
horizon=4L,
outcome="flu.admits",
trim_date=NULL,
covariates=TRUE)
# Format for submission
hosp_formatted <- ts_format_for_submission(hosp_fitfor$tsfor)
# Plot with current and all data
plot_forecast(prepped_hosp, hosp_formatted$ensemble)
plot_forecast(prepped_hosp_all, hosp_formatted$ensemble)
plot_forecast(prepped_hosp, hosp_formatted$ensemble, location=c("US", "51"))
plot_forecast(prepped_hosp_all, hosp_formatted$ensemble, location=c("US", "51"))
plot_forecast(prepped_hosp, hosp_formatted$ets)
plot_forecast(prepped_hosp_all, hosp_formatted$ets)
plot_forecast(prepped_hosp, hosp_formatted$arima)
plot_forecast(prepped_hosp_all, hosp_formatted$arima)
# Demonstrating multiple models
prepped_hosp <-
h_raw %>%
prep_hdgov_hosp(statesonly=TRUE, min_per_week = 0, remove_incomplete = TRUE) %>%
dplyr::filter(abbreviation != "DC") %>%
dplyr::filter(week_start < as.Date("2022-01-08", format = "%Y-%m-%d"))
tsens_20220110 <-
system.file("extdata/2022-01-10-SigSci-TSENS.csv", package="fiphde") %>%
readr::read_csv(show_col_types = FALSE)
creg_20220110 <-
system.file("extdata/2022-01-10-SigSci-CREG.csv", package="fiphde") %>%
readr::read_csv(show_col_types = FALSE)
combo_20220110 <- dplyr::bind_rows(
dplyr::mutate(tsens_20220110, model = "SigSci-TSENS"),
dplyr::mutate(creg_20220110, model = "SigSci-CREG")
)
plot_forecast(prepped_hosp, combo_20220110, location = "24")
plot_forecast(prepped_hosp, tsens_20220110, location = "24")
plot_forecast(prepped_hosp, combo_20220110, location = c("34","36"))
plot_forecast(prepped_hosp, creg_20220110, location = "US", .model = "SigSci-CREG")
plot_forecast(prepped_hosp, creg_20220110, location = "US", .model = "SigSci-CREG")
## demonstrating different prediction interval widths
plot_forecast(prepped_hosp, combo_20220110, location = "24", pi = 0.5)
plot_forecast(prepped_hosp, combo_20220110, location = "24", pi = 0.9)
plot_forecast(prepped_hosp, combo_20220110, location = "24", pi = 0.95)
plot_forecast(prepped_hosp, combo_20220110, location = "24", pi = NULL)
} # }