logScoreMultDirIncremental: Multinomial-Dirichlet Log marginal likelihood (online).

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

Description

Compute the difference in log marginal likelihood of the supplied Bayesian Networks, quickly.

Usage

  logScoreMultDirIncremental(currentBN, proposalBN, heads,
    logScoreParameters, cache, checkInput = T)

Arguments

currentBN	An object of class "bn".
proposalBN	An object of class "bn".
heads	A numeric vector, specifying which nodes have different parents in currentBN and proposalBN.
logScoreParameters	A list with the following components: data A matrix (NOT data.frame), with columns being integers in the range 0, 1, 2, .... giving the values of each random variable. **** The integers MUST start numbering at 0 NOT 1 **** nl A numeric vector of length ncol(data), specifying the number of levels that each random variable takes. hyperparameters A character vector of length one. Either "bdeu", "qi", "one", or "point9"
cache	Optionally, provide an environment with cached local scores for this data.
checkInput	A logical of length 1, specifying whether to check the inputs to the function.

Details

This is a fast, incremental version of logScoreMultDir.

*** NOTE: THIS REQUIRES A MATRIX RATHER THAN A DATA FRAME ***

The data must be discrete. The conditional distributions of each random variable, conditional on its parents are assumed to be multinomial, with Dirichlet priors for the parameters.

The notation here roughly follows Mukherjee and Speed (2008) Network inference using informative priors. PNAS 105 (38) 14313-14318, doi: 10.1073/pnas.0802272105

Value

logscore(proposalBN) - logscore(currentBN)

See Also

logScoreMultDir, logScoreMultDirOffline

Examples

x1 <- factor(c(1, 1, 0, 1, 0, 0, 1, 0, 1, 0))
x2 <- factor(c(0, 1, 0, 1, 0, 1, 1, 0, 1, 0))
x3 <- factor(c(0, 1, 1, 1, 0, 1, 1, 0, 1, 0))
data <- data.frame(x1 = x1, x2 = x2,  x3 = x3)

n1 <- bn(numeric(0), 1, 2)
n2 <- bn(numeric(0), 1, numeric(0))

logScoreMultDir(n1, data)
logScoreMultDir(n2, data)

data <- data.frame(lapply(data, as.factor))
data <- data.matrix(data) - 1
cc <- new.env(hash = TRUE, size = 10000L)
nl <- apply(data, 2, function(i) length(unique(i)))
names(nl) <- seq_along(n1)
lsp <- list(data = data, nl = nl, hyperparameters = "qi")
logScoreMultDirIncremental(n1, n2, 3, lsp, cc)

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