ergmm() is used to fit latent space and latent space cluster
random network models, as described by Hoff, Raftery and Handcock (2002),
Handcock, Raftery and Tantrum (2005), and Krivitsky, Handcock, Raftery, and
Hoff (2009). ergmm() can return either a Bayesian model fit or
the two-stage MLE.
ergmm(
formula,
response = NULL,
family = "Bernoulli",
fam.par = NULL,
control = control.ergmm(),
user.start = list(),
prior = ergmm.prior(),
tofit = c("mcmc", "mkl", "mkl.mbc", "procrustes", "klswitch"),
Z.ref = NULL,
Z.K.ref = NULL,
seed = NULL,
verbose = FALSE
)ergmm returns an object of class
ergmm containing the information about the
posterior.
An formula object, of the form g ~ <term 1> +
<term 2> ..., where g is a network object or a matrix
that can be coerced to a network object, and <term 1>,
<term 2>, etc., are each terms for the model. See
ergmTerm for the terms that can be fitted, though note the
section on fixed effects below. To create a network object in ,
use the network function, then add nodal attributes to it
using set.vertex.attribute if necessary.
Note that, as in lm(), the model will include an
intercept term. This behavior can be overridden by
including a -1 or +0 term in the formula, and a
1(mean=...,var=...) term can be used
to set a prior different from default.
An optional edge attribute that serves as the response
variable. By default, presence (1) or absence (0) of an edge in g is
used.
A character vector specifying the conditional distribution of each edge value. See families.ergmm for the currently implemented families.
For those families that require additional parameters, a list.
The MCMC parameters that do not affect the posterior
distribution such as the sample size, the proposal variances, and tuning
parameters, in the form of a named list. See control.ergmm for
more information and defaults.
An optional initial configuration parameters for MCMC in the form of a list. By default, posterior mode conditioned on cluster assignments is used. It is permitted to only supply some of the parameters of a configuration. If this is done, the remaining paramters are fitted conditional on those supplied.
The prior parameters for the model being fitted in the form of
a named list. See term help for the terms to use. If given, will
override those given in the formula terms, making it useful as a convenient
way to store and reproduce a prior distribution. The list or prior
parameters can also be extracted from an ERGMM fit
object. See ergmm.prior for more information.
A character vector listing some subset of "pmode", "mcmc",
"mkl", "mkl.mbc", "mle","procrustes", and "klswitch", defaulting to all of
the above, instructing ergmm what should be returned as a part
of the ERGMM fit object. Omiting can be used to skip
particular steps in the fitting process. If the requested procedure or
output depends on some other procedure or output not explictly listed, the
dependency will be resolved automatically.
If given, used as a reference for Procrustes analysis.
If given, used as a reference for label-switching.
If supplied, random number seed.
If this is TRUE (or 1), causes information to
be printed out about the progress of the fitting, particularly initial value
generation. Higher values lead to greater verbosity.
Each coefficient for a fixed effect covariate has a normal prior whose
mean and variance are set by the mean and var parameters
of the term. For those formula terms that add more than one covariate,
a vector can be given for mean and variance. If not, the vectors given
will be repeated until the needed length is reached.
ergmm can use model terms implemented for the
ergm package and via the
ergm.userterms
API (in GitHub repository statnet/ergm.userterms). See
ergmTerm for a list of available terms. If you
wish to specify the prior mean and variance, you can add them to
the call. E.g.,
TERMNAME(..., mean=0, var=9),
where
... are the arguments for the ergm term, will
initialize TERMNAME with prior mean of 0 and prior variance
of 9.
Some caveats:
ergm has a binary and a valued
mode. Regardless of the family
used, the binary variant of the
ergm term will be used in the
linear predictor of the model.
ergm does not support modeling
self-loops, so terms imported in this way will always have
predictor x[i,i]==0. This should not affect most
situations, but if you absolutely must model self-loops and
non-self-edges in one term, use the deprecated terms below.
latentnet only fits models with dyadic
independence. Terms that induce dyadic dependence (e.g.,
triangles) can be used, but then the likelihood of the
model will, effectively, be replaced with
pseudolikelihood. (Note that under dyadic independence, the two
are equal.)
Mark S. Handcock, Adrian E. Raftery and Jeremy Tantrum (2002). Model-Based Clustering for Social Networks. Journal of the Royal Statistical Society: Series A, 170(2), 301-354.
Peter D. Hoff, Adrian E. Raftery and Mark S. Handcock (2002). Latent space approaches to social network analysis. Journal of the American Statistical Association, 97(460), 1090-1098.
Pavel N. Krivitsky, Mark S. Handcock, Adrian E. Raftery, and Peter D. Hoff (2009). Representing degree distributions, clustering, and homophily in social networks with latent cluster random effects models. Social Networks, 31(3), 204-213.
Pavel N. Krivitsky and Mark S. Handcock (2008). Fitting Position
Latent Cluster Models for Social Networks with latentnet. Journal of
Statistical Software, 24(5). tools:::Rd_expr_doi("10.18637/jss.v024.i05")
network, summary.ergmm, ergmTerm, families.ergmm
# \donttest{
#
# Use 'data(package = "latentnet")' to list the data sets in a
#
data(package="latentnet")
#
# Using Sampson's Monk data, lets fit a
# simple latent position model
#
data(sampson)
samp.fit <- ergmm(samplike ~ euclidean(d=2))
#
# See if we have convergence in the MCMC
mcmc.diagnostics(samp.fit)
#
# Plot the fit
#
plot(samp.fit)
#
# Using Sampson's Monk data, lets fit a latent clustering random effects model
#
samp.fit2 <- ergmm(samplike ~ euclidean(d=2, G=3)+rreceiver)
#
# See if we have convergence in the MCMC
mcmc.diagnostics(samp.fit2)
#
# Plot the fit.
#
plot(samp.fit2, pie=TRUE)
# }
Run the code above in your browser using DataLab