variance_emulator: Variance Emulator builder
In Tandethsquire/emulatorr: Emulation and History Matching Package

Creates an emulator whose variance itself is emulated, for stochastic systems.

variance_emulator(
  input_data_var,
  input_data_exp,
  npoints,
  output_names,
  ranges,
  kurt = 3,
  input_names = names(ranges),
  beta,
  u,
  c_lengths,
  funcs,
  bucov,
  deltas,
  ev,
  quadratic = TRUE,
  beta.var = FALSE,
  lik.method = "my"
)

`input_data_var`	Required. A `data.frame` containing the input parameters and output variances from a set of simulator runs.
`input_data_exp`	Required. A `data.frame` containing the input parameters and output means from a set of simulator runs.
`npoints`	The number of runs performed per observed point.
`output_names`	Required. The list of outputs to emulate from `input_data`.
`ranges`	A named list of parameter ranges.
`kurt`	The value of the kurtosis. Default: 3
`input_names`	A list of input_names (if `ranges` is not provided).
`beta`	Optional: specifications for the regression coefficients, given as a list of lists `list(mu, sigma)` (a la `Emulator` specification).
`u`	Optional: the correlation structure for each output, given as a list of lists `list(mu, sigma, corr)`.
`c_lengths`	Optional: a set of correlation lengths.
`funcs`	Optional: basis functions for the regression surface.
`bucov`	Optional: a list of functions giving the covariance between each of the beta parameters and u(x).
`deltas`	Optional: the nugget terms to include in u(x).
`ev`	Optional. Used for determining nugget terms in absence on delta
`quadratic`	Optional: should the regression surface be linear or quadratic? Default: F
`beta.var`	Optional: should the beta coefficient be assumed to be known or should model variance be included?
`lik.method`	Optional: method used to determine hyperparameters sigma and theta.