Make a Forecaster

A forecaster is an object that can make forecasts. It is constructed from mp_tmb_calibrator, which is an object that can be used to calibrate the parameters of a model – typically by fitting the model to data. The main difference between a calibrator and a forecaster is that the latter runs for more time steps, past the last time step of the calibrator. A forecaster object can be used to generate different types of forecasts using the mp_trajectory family of functions.

Usage

mp_forecaster(
  calibrator,
  forecast_period_time_steps,
  outputs = NULL,
  data = NULL,
  tv = NULL,
  default = list()
)

Arguments

calibrator

Object made using mp_tmb_calibrator, which can be calibrated to data.

forecast_period_time_steps

The number of time steps to project beyond the period with data.

outputs

An optional character vector of variables in the model to forecast. If this argument is omited, then the forecaster will inherit the outputs of the calibrator object. Note that if you prefix the name of an output using log_, logit_, or sqrt_ then the transformed version of the outputs will be simulated. This technique is useful for confidence intervals produced by mp_trajectory_sd that go out of the valid range of values (e.g., use log_ for confidence intervals of negative state variables).

data

An optional data frame containing the data that were fitted to. Typically this argument will be unused, because by default the data are stored in the calibrator unless you want to avoid making too many copies of the data.

tv

An optional replacement for the tv parameter in the mp_tmb_calibrator function.

default

An optional list of default model variables (e.g., parameters initial values of state variables) to override calibrated values.