residualsMultDir: Residuals from a Multinomial-Dirichlet model

In rjbgoudie/structmcmc: Structural inference for Bayesian Networks and Variable Selection.

Description

Given a Bayesian network, some training data and some test data, the model given by fitting the Bayesian network to the training data is used to predict each node of the test data, given the parents of that node in the Bayesian network.

Usage

  residualsMultDir(x, weights = 1, train, test,
    metric = kronecker_delta, verbose = F)

Arguments

x	A BN, or a bn.list
weights	A numeric vector weights for the models x
train	A data frame of training data
test	A data frame of test data
metric	A function that measures the distance between the predictions and the true values
verbose	Logical indicating whether verbose output should be given

Details

The residual is then computed, using the supplied metric.

Alternatively, a bn.list of Bayesian networks can be supplied, together with a vector of weights. The models (Bayesian networks) are then averaged over, according to the supplied weights, to give a model averaging prediction.

The residuals are again computed, using the suppplied metric.

rjbgoudie/structmcmc documentation built on Nov. 3, 2020, 3:41 a.m.