dga: Fit decomposable graphical models

In OlivierBinette/MSETools: Interface to Multiple Systems Estimation Software

Fit decomposable graphical models

Description

Fit the decomposable graphical model approach of Madigan and York (1997), as implemented in the dga package of Kristian Lum, James Johndrow, and Patrick Ball (2015).

Usage

dga(data, priorCounts = NULL, graphs = NULL, Nmissing = NULL,
 logPriorGraphs = NULL, logPriorN = NULL, maxMissingMult = 30,
 length.out = 1000)

Arguments

data	object of class "MSEdata" representing list inclusion pattern counts for between 3 and 5 lists.
priorCounts	"Prior counts" for the hyper-Dirichlet prior on decomposable graphs. This can be a positive numeric value for constant counts or an L-dimensional array (where L is the number of lists) representing the table of prior counts (see examples below).
graphs	List of decomposable graphs under consideration. Must be a sub-list of "graphs3", "graphs4" or "graphs5". Default is "graphsL" where L is the number of lists (between 3 and 5).
Nmissing	integer vector for the number of plausible unobserved individuals. Population size posterior probability is computed for the values corresponding to these numbers of missing individuals. Defaults to a vector of length 1000 ranging between 1 and 30 times the number of observed individuals.
logPriorGraphs	Numeric vector of log prior probabilities over the set of decomposable graphs in "graphs". Defaults to a constant prior.
logPriorN	Numeric vector of log prior probabilities of the number of missing individuals in "Nmissing". Defaults to the population size prior $p(N) \propto 1/N$. Nmissing should be specified explicitely if logPriorN is provided.
maxMissingMult	If Nmissing is NULL, then maxMissingMult determines the default range of Nmissing: between 1 and maxMissingMult times the number of observed individuals.
length.out	If Nmissing is NULL, then length.out determines the maximum default length of Nmissing.

See Also

estimates MSEdata

Examples

dga_fit <- dga(UK)
estimates(dga_fit)

OlivierBinette/MSETools documentation built on Aug. 7, 2022, 8:42 p.m.