ergm.bridge.llr {ergm}R Documentation

A simple implementation of bridge sampling to evaluate log-likelihood-ratio between two ERGM configurations

Description

This function uses bridge sampling with geometric spacing to estimate the difference between the log-likelihoods of two parameter configurations for an ERGM via repeated calls to simulate.formula.ergm.

Usage

ergm.bridge.llr(object, from, to, nsteps = 20,
sample.size = 10000, burnin = 10000, basis = NULL, verbose = FALSE, llronly = FALSE, ...)

Arguments

object

A model formula. See ergm for details.

basis

See simulate.ergm.

verbose

How much information to print about the progress.

from

The initial parameter configuration.

to

The final parameter configuration.

nsteps

Number of geometric bridges to use.

sample.size

Total number of MCMC draws to use (to be divided up among the bridges, so each bridge gets sample.size/nsteps draws.

burnin

Total number of steps draws to discart for each bridge (to be divided up among the bridges, so each bridge after the first gets sample.size/sqrt(nsteps).

llronly

Whether only the estiamted log-ratio should be returned. (Defaults to FALSE.)

...

Further arguments to simulate.formula.ergm.

Value

If llronly=TRUE, returns the scalar log-likelihood-ratio. Otherwise, returns a list with the following components:

llr

The estimated log-ratio.

llrs

The estimated log-ratios for each of the nsteps bridges.

path

A numeric matrix with nsteps rows, with each row being the respective bridge's parameter configuration.

stats

A numeric matrix with nsteps rows, with each row being the respective bridge's vector of simulated statistics.

Dtheta.Du

The gradient vector of the parameter values with respect to position of the bridge.

References

Hunter, D. R. and Handcock, M. S. (2006) Inference in curved exponential family models for networks, Journal of Computational and Graphical Statistics.

See Also

simulate.formula.ergm


[Package ergm version 2.4-3 Index]