make_node_states_estimates: Create 'estimate' for Bayesain network node

In Issoufou-Liman/decisionSupportExtra: Miscellaneous Functions for Quantitative Support of Decision Making under Uncertainty

Description

create a multivariate estimate object for Bayesian network node. The function samples from the posterior distribution of the Bayesian network to construct an object of class estimate accounting for each state of the node.

Usage

make_node_states_estimates(
  bn,
  node,
  op = "proba",
  distr = "beta",
  n_run = 1000,
  state_effects = NULL,
  evidence = NULL,
  n_generation = NULL,
  include_relatives = TRUE,
  estimate_method = "fit",
  percentiles = c(0.025, 0.5, 0.975),
  plot = FALSE,
  show.output = FALSE,
  fit = c("emp", "theo"),
  emp_estimates = NULL
)

Arguments

bn	an object of class bn.fit.
node	character string, the label of the node which conditional distribution is of interest.
op	a vector of character strings, the type of returned value: either probabilities or raw data sampled from the posterior distribution
distr	charater vector. The expected probability distribution function.
n_run	integer specifying the number of of model run. Default is 1000.
state_effects	numeric vector specifying the relative weight factor of each state in the final estimate.
evidence	a name value pair:a named character vector which values are node states and names are node names.
n_generation	how far to go in the network topology for building the conditionning specification for the query?
include_relatives	logical Should parents or ancestors, depending on the the argument n_generation, should be included in the query? If TRUE, the default, these will be internally involved in constructing the evidence argument.
estimate_method	character vector. Method to be used for estimating the marginal distribution parameters as in estimate. If null, the default, method is set to "fit" (see estimate).
percentiles	numeric vector of percentiles based on which to define a quantile function to be used for fitting distr with rriskDistributions package.
plot	logical. Should the quantile function be ploted? default to FALSE
show.output	Should the computation be verbose? default to FALSE
fit	character specifying whether the estimates should be derived from either empirical ('emp', default) or theoritical distribution ('theo').
emp_estimates	list alternative estimates specification.Default to NULL. This is only used when fit = 'theo'.

Details

see estimate.

Author(s)

Issoufou Liman

References

Eike Luedeling and Lutz Goehring (2018). decisionSupport: Quantitative Support of Decision Making under Uncertainty. R package version 1.103.8. https://CRAN.R-project.org/package=decisionSupport

See Also

estimate.

Examples

library (gRain)
library(bnlearn)
## setting a bayesian network with gRain
Soil_type <- cptable (~Soil_type, values = c(0.05, 0.55, 0.4),
levels = c('Sandy', 'Loamy', 'Clayey'))
Manure_application <- cptable(~Manure_application, values = c(0.3, 0.7),
levels = c('FALSE', 'TRUE'))
Soil_water_holding_capacity_tmp <- make_gRain_CPT(
 parent_effects = list(c(0, 2.5, 3), c(0, 2)),
 parent_weights = c(2,1),
 b = 3,
 child_prior = c(0.2,0.5,0.3),
 child_states = c('Low', 'Medium', 'High'),
 parent_states = list(c('Sandy', 'Loamy', 'Clayey'), c('FALSE', 'TRUE'))
)
Soil_water_holding_capacity_values <- Soil_water_holding_capacity_tmp$values
Soil_water_holding_capacity_levels <- Soil_water_holding_capacity_tmp$levels
Soil_water_holding_capacity <- cptable (
~Soil_water_holding_capacity|Soil_type:Manure_application,
values = Soil_water_holding_capacity_values,
levels = Soil_water_holding_capacity_levels)
## Compile conditional probability tables
network <- compileCPT(list(Soil_type, Manure_application, Soil_water_holding_capacity))
## Graphical Independence Network ####
network <- grain(network)
## Use grain object (gRain package)
make_node_states_estimates (bn = network, node = 'Soil_water_holding_capacity')

## converting the grain bayesian network to bn.fit
network_bn_fit <- as.bn.fit(network)
## Use bn.fit object (bnlearn package)
make_node_states_estimates (bn = network_bn_fit,
node = 'Soil_water_holding_capacity', distr = c('beta', 'norm', 'gamma'))

Issoufou-Liman/decisionSupportExtra documentation built on Dec. 21, 2020, 6:28 p.m.