This function is a helper that forecasts Poisson counts for near-term horizons based on characteristics of recently observed count data. The function effectively takes a rolling average of most recent observations (augmenting with each forecasted horizon as the horizons progress), then uses this average as the parameter for Lambda in a random draw from a Poisson distribution.
Arguments
- .data
Data frame with incoming data that includes a variable with counts (see ".var" argument), and location (must be stored in a column called "location") and a variable for sorting by date (must be stored in a column called "week_start")
- .location
The name of the location of interest
- .var
Bare, unquoted name of the variable with counts to be forecasted
- horizon
The number of horizons ahead to forecast; must be one of
4or5; default is4
Examples
if (FALSE) { # \dontrun{
all_clin <- get_cdc_clin()
va_ahead <-
dplyr::tibble(
n_positive = pois_forc(all_clin, .location = "51", n_positive),
total = pois_forc(all_clin, .location = "51", total),
p_positive = n_positive / total)
va_ahead
} # }