HierarchicalEmulator {hmer}

Hierarchical Bayes Linear Emulator

Description

Creates a univariate emulator with hierarchical structure.

This object does not differ extensively from the standard Emulator object, so most of the functionality will not be listed here: the main difference is that it allows for the variance structure of the emulator to be modified by a higher order object. The typical usage is to create a variance emulator, whose predictions inform the behaviour of a mean emulator with regard to a stochastic process.

Constructor

HierarchicalEmulator$new(basis_f, beta, u, ranges, ...)

Arguments

For details of shared arguments, see Emulator.

s_diag The function that modifies the structure of the Bayes Linear adjustment.

samples A numeric vector that indicates how many replicates each of the training points has.

em_type Whether the emulator is emulating a mean surface or a variance surface.

Constructor Details

See Emulator: the constructor structure is the same save for the new arguments discussed above.

Accessor Methods

get_exp(x, samps = NULL) Similar in form to the normal Emulator method; the samps argument allows the estimation of summary statistics derived from multiple realisations.

get_cov(x, xp = NULL, full = FALSE, samps = NULL) Differences here are in line with those described in get_exp.

Object Methods

Identical to those of Emulator: the one internal difference is that adjust returns a HierarchicalEmulator rather than a standard one.

References

Goldstein & Vernon (2016), in preparation

Examples

 h_em <- emulator_from_data(BirthDeath$training, c('Y'),
  list(lambda = c(0, 0.08), mu = c(0.04, 0.13)), emulator_type = "variance")
 names(h_em) # c("expectation', 'variance')