variance_emulator_from_data {hmer}

Variance Emulator Creation (Deprecated)

Description

Trains hierarchical emulators to stochastic systems

Usage

variance_emulator_from_data(
  input_data,
  output_names,
  ranges,
  input_names = names(ranges),
  verbose = interactive(),
  na.rm = FALSE,
  ...
)

Arguments

input_data

All model runs at all points.

output_names

The observation names.

ranges

A named list of parameter ranges

input_names

The names of the parameters (if ranges is not provided).

verbose

Should status updates be printed to console?

na.rm

Should NA values be removed before training?

...

Optional parameters that can be passed to link{emulator_from_data}.

Details

This function is deprecated in favour of using emulator_from_data with argument emulator_type = "variance". See the associated help file.

For stochastic systems, one may emulate the variance as well as the function itself. This is particularly true if one expects the variance to be very different in different areas of the parameter space (for example, in an epidemic model). This function performs the requisite two-stage Bayes Linear update.

All observations are required (including replicates at points) - this function collects them into the required chunks and calculates the summary statistics as required.

All other parameters passed to this function are equivalent to those in emulators are the Bayes Linear adjusted forms.

Value

A list of lists: one for the variance emulators and one for the function emulators.

References

Goldstein & Vernon (2016) in preparation

Examples

 # Excessive runtime
 # A simple example using the BirthDeath dataset
 v_ems <- variance_emulator_from_data(BirthDeath$training, c("Y"),
  list(lambda = c(0, 0.08), mu = c(0.04, 0.13)), c_lengths = c(0.75))