full_wave: History Match
In Tandethsquire/emulatorr: Emulation and History Matching Package

Description Usage Arguments Details Value Examples

Performs a full wave of emulation and history matching from data.

full_wave(
  input_data,
  validation_data,
  ranges,
  output_names,
  targets,
  n_points = 40,
  previous_wave = NULL,
  sample_method = "importance",
  ...
)

`input_data`	The set of training points
`validation_data`	The set of points to use in validation
`ranges`	The ranges of the inputs, as a named list.
`output_names`	The names of the outputs to emulate.
`targets`	The observations, given in the usual `(val, sigma)` form
`n_points`	The number of points to evaluate the parameters on.
`previous_wave`	The preliminary emulators for the set of waves, if they exist
`sample_method`	The method to be used to find new points (see `generate_new_runs`)
`...`	Any optional parameters to pass to `emulator_from_data`

Given simulator runs (split into training and validation data), the target values, and the identification of outputs to emulate, the function generates trained emulators, tests them with emulator diagnostics (removing any emulators whose outputs cannot be well emulated from the data), and finally generates a new sample of points to be entered into the simulator.

Necessary parameters to be passed are the input data, the validation data, the ranges of inputs, and the observation values for each output. If any specifications are to be passed directly to the emulator construction, then they should be given as additional parameters (see emulator_from_data to see the options).

A set of preliminary ('wave 0') emulators are fitted to the data before being used to train a new set of emulators on the data, using Bayes linear adjustment. The preliminary emulators are provided as part of the output of the function to indicate the prior specifications, should any by-hand modification be needed (for example, if any of the outputs could not be adequately fitted).

The output consists of a list of four items: the preliminary emulators base_emulators, the trained emulators emulators, the next points to be put into the simulator next_sample, and the minimum enclosing hyperrectangle for the non-implausible region, given as ranges new_ranges.

A list of base emulators, trained emulators for this wave, new sample points, and new ranges.

 #ranges <- list(aSI = c(0.1, 0.8), aIR = c(0, 0.5), aSR = c(0, 0.05))
 #outputs <- c('nS','nI','nR')
 #targets <- list(
 # list(val = 281, sigma = 10.43),
 # list(val = 30, sigma = 11.16),
 # list(val = 689, sigma = 14.32)
 #)
 #wave1 <- full_wave(GillespieSIR, GillespieValidation, ranges, outputs, targets,
 # n_points = 30, deltas = rep(0.1, 3), quadratic = TRUE)