add_response_columns: Compute the corresponding observed response values for a set...

In dshemetov/modeltools-mirror: Tools for Building COVID Forecasters and Related Models

Description

The response for an observed data point at time t is defined as the mean of observed values on days t + s through t + e (inclusive) multiplied by e - s, where s and e are the number of days from forecast_date to the start and end, respectively, of the next full incidence_period. We choose to take the mean and multiply by the incidence period length rather than simply sum the observed values in order to avoid treating missing values as 0.

Usage

add_response_columns(
  signals,
  response_name,
  forecast_date,
  incidence_period,
  ahead
)

Arguments

signals	Wide data frame of signals. We expect the following columns: geo_valueStrings of geographic locations. time_valueDates of training data. Covariate columnsColumns with names of the form value-{days}:{signal} or value+0:{signal} whose values correspond to signaldaysbeforetime_value'
response_name	Name of response signal. The value of the response variable on a single day should be located in column value+0:{response_name} of signals.
forecast_date	Date on which the forecast will be made
incidence_period	Time period over which the response should be summed.
ahead	integer or vector of integer ahead values

Value

Data frame of signals equal to signals with length(ahead) additional columns with names response+{a}:{response_name} corresponding to the response variable at ahead a.

Examples

## Not run: 
add_response_columns(
  tibble(
    geo_value = "mi"
    time_value = seq.Date(as.Date("2020-07-01"), as.Date("2020-07-14"), "day"),
    `value+0:sig` = 1:14
  ),
  "sig",
  as.Date("2020-07-15"),
  "epiweek",
  1:2 
)

## End(Not run)

dshemetov/modeltools-mirror documentation built on Jan. 7, 2022, 12:23 a.m.