| Title: | Fit, Simulate and Diagnose Models for Network Evolution Based on Exponential-Family Random Graph Models |
|---|---|
| Description: | An integrated set of extensions to the 'ergm' package to analyze and simulate network evolution based on exponential-family random graph models (ERGM). 'tergm' is a part of the 'statnet' suite of packages for network analysis. See Krivitsky and Handcock (2014) <doi:10.1111/rssb.12014> and Carnegie, Krivitsky, Hunter, and Goodreau (2015) <doi:10.1080/10618600.2014.903087>. |
| Authors: | Pavel N. Krivitsky [aut, cre]  ,
Mark S. Handcock [aut, ths],
David R. Hunter [ctb],
Steven M. Goodreau [ctb, ths],
Martina Morris [ctb, ths],
Nicole Bohme Carnegie [ctb],
Carter T. Butts [ctb],
Ayn Leslie-Cook [ctb],
Skye Bender-deMoll [ctb],
Li Wang [ctb],
Kirk Li [ctb],
Chad Klumb [ctb],
Adrien Le Guillou [ctb] |
| Maintainer: | Pavel N. Krivitsky <[email protected]> |
| License: | GPL-3 + file LICENSE |
| Version: | 4.2-2564 |
| Built: | 2024-11-05 04:21:51 UTC |
| Source: | https://github.com/statnet/tergm |
An integrated set of extensions to the 'ergm' package to analyze and simulate network evolution based on exponential-family random graph models (ERGM). 'tergm' is a part of the 'statnet' suite of packages for network analysis. See Krivitsky and Handcock (2014) doi:10.1111/rssb.12014 and Carnegie, Krivitsky, Hunter, and Goodreau (2015) doi:10.1080/10618600.2014.903087.
tergm is a collection of extensions to the
ergm package to fit, diagnose, and simulate
models for dynamic networks — networks that evolve over time — based on
exponential-family random graph models (ERGMs). For a list of functions type
help(package='tergm')
When publishing results obtained using this package, please cite the
original authors as described in citation(package="tergm").
All programs derived from this package must cite it.
An exponential-family random graph model (ERGM) postulates an exponential family over the sample space of networks of interest, and ergm package implements a suite of tools for modeling single networks using ERGMs.
There have been a number of extensions of ERGMs for modeling the evolution of networks, including the temporal ERGM (TERGM) of Hanneke et al. (2010) and the separable termporal ERGM (STERGM) of Krivitsky and Handcock (2014). The latter model allows familiar ERGM terms and statistics to be reused in a dynamic context, interpreted in terms of formation and dissolution (persistence) of ties. Krivitsky (2012) suggested a method for fitting dynamic models when only a cross-sectional network is available, provided some temporal information for it is available as well.
This package aims to implement these and other ERGM-based models for network
evolution. At this time, it implements, via the tergm()
function, a general framework for modeling tie dynamics in temporal networks
with flexible model specification (including (S)TERGMs). Estimation options
include a conditional MLE (CMLE) approach for fitting to a series of
networks and an Equilibrium Generalized Method of Moments Estimation (EGMME)
for fitting to a single network with temporal information. For further
development, see the referenced papers.
The operator terms implemented by tergm are Form(),
Persist(), Diss(), Cross(), and Change(). These are used
to specify how the ergm terms (ergmTerm) in a formula are
evaluated across a network time-series. Note, you cannot use one
of these operators within another temporal, so
Cross(~Form(~edges)) is not a valid specification. (Generally,
nesting these operators within other operators will often not work;
nesting other operators within them will almost always work,
however.)
The durational terms are distinguished either by their name,
mean.age, or their name extensions: <name>.ages,
<name>.mean.age, and <name>.age.interval. In contrast to
their eponymous terms in ergm, these durational
terms take into account the elapsed time since each (term-relevant) dyad in
the network was last toggled.
As currently implemented, the package does not support use of many durational terms during estimation, though it may work with some. But durational terms may be used as targets, monitors, or summary statistics. The ability to use these terms in the estimation of models is under development.
If you previously used the stergm() function in this package, please
note that stergm() has been superceded by the new tergm()
function, and has been deprecated. The
dissolution formula in stergm() maps to the new Persist()
operator in the tergm() function, not the Diss() operator.
For detailed information on how to download and install the software, go to the Statnet project website: https://statnet.org. A tutorial, support newsgroup, references and links to further resources are provided there.
Maintainer: Pavel N. Krivitsky [email protected] (ORCID)
Authors:
Mark S. Handcock [email protected] [thesis advisor]
Other contributors:
David R. Hunter [email protected] [contributor]
Steven M. Goodreau [email protected] [contributor, thesis advisor]
Martina Morris [email protected] [contributor, thesis advisor]
Nicole Bohme Carnegie [email protected] [contributor]
Carter T. Butts [email protected] [contributor]
Ayn Leslie-Cook [email protected] [contributor]
Skye Bender-deMoll [email protected] [contributor]
Li Wang [email protected] [contributor]
Kirk Li [email protected] [contributor]
Chad Klumb [email protected] [contributor]
Adrien Le Guillou [email protected] (ORCID) [contributor]
Hanneke S, Fu W and Xing EP (2010). Discrete Temporal Models of Social Networks. Electronic Journal of Statistics, 2010, 4, 585-605. doi:10.1214/09-EJS548
Krackhardt, D and Handcock, MS (2006) Heider vs Simmel: Emergent features in dynamic structures. ICML Workshop on Statistical Network Analysis. Springer, Berlin, Heidelberg, 2006.
Krivitsky PN & Handcock MS (2014) A Separable Model for Dynamic Networks. Journal of the Royal Statistical Society, Series B, 76(1): 29-46. doi:10.1111/rssb.12014
Krivitsky, PN (2012). Modeling of Dynamic Networks based on Egocentric Data with Durational Information. Pennsylvania State University Department of Statistics Technical Report, 2012(2012-01). https://web.archive.org/web/20170830053722/https://stat.psu.edu/research/technical-report-files/2012-technical-reports/TR1201A.pdf
Butts CT (2008). network: A Package for Managing Relational Data in . Journal of Statistical Software, 24(2). doi:10.18637/jss.v024.i02
Goodreau SM, Handcock MS, Hunter DR, Butts CT, Morris M (2008a). A statnet Tutorial. Journal of Statistical Software, 24(8). doi:10.18637/jss.v024.i08
Hunter, D. R. and Handcock, M. S. (2006) Inference in curved exponential family models for networks, Journal of Computational and Graphical Statistics, 15: 565-583
Hunter DR, Handcock MS, Butts CT, Goodreau SM, Morris M (2008b). ergm: A Package to Fit, Simulate and Diagnose Exponential-Family Models for Networks. Journal of Statistical Software, 24(3). doi:10.18637/jss.v024.i03
Morris M, Handcock MS, Hunter DR (2008). Specification of Exponential-Family Random Graph Models: Terms and Computational Aspects. Journal of Statistical Software, 24(4). doi:10.18637/jss.v024.i04
Useful links:
Report bugs at https://github.com/statnet/tergm/issues
This function is used in tergm.EGMME.initialfit and also when targets or monitoring
formulas are specified by characters. It makes a basic attempt to identify the
formation and dissolution formulas within a larger combined formula (which may also
include non-separable terms). Instances of Form at the top level (which may occur
inside offset) contribute to the formation formula; instances of Persist and
Diss at the top level (which may also occur inside offset) contribute to the
dissolution formula. All other terms are regarded as non-separable; this includes instances
of Form, Persist, and Diss that occur inside other operator terms,
including inside Offset, and also includes all interactions at the top level (for which
the top level term is effectively the interaction operator * or :),
whether or not they include Form, Persist, and/or Diss.
The formation and dissolution formulas are obtained by adding
the contributing terms, replacing Form and Persist with trivial operators that protect
the environments of their formula arguments but have no effect on statistics or coefficient names
(meaning the formulas effectively become cross-sectional), and replacing Diss by a similar operator
that negates statistics. These are included in the return value as the form and pers
elements of the list (the "dissolution" formula really being the persistence formula), which also includes
the formula of non-separable terms as nonsep, and the formula of all terms after replacing
Form, Persist, and Diss as described above as all.
If usage proves problematic, one may specify the monitoring and/or targets formulas explicitly
(rather than by characters), and one may pass initial coefficient values for the EGMME to avoid
running tergm.EGMME.initialfit.
.extract.fd.formulae(formula).extract.fd.formulae(formula)
formula |
a |
A list containing form, pers, nonsep, and all formulas as described above.
The Change Operator Term
# binary: Change( # formula, # lm = ~1, # subset = TRUE, # weights = 1, # contrasts = NULL, # offset = 0, # label = NULL # )# binary: Change( # formula, # lm = ~1, # subset = TRUE, # weights = 1, # contrasts = NULL, # offset = 0, # label = NULL # )
formula |
a one-sided |
lm, subset, weights, contrasts, offset, label
|
|
This term accepts a model formula and produces the corresponding model for a network constructed by taking the dyads that have changed between time steps.
ergmTerm for index of model terms currently visible to the package.
None
Auxiliary function as user interface for fine-tuning STERGM simulation.
control.simulate.network( MCMC.burnin.min = 1000, MCMC.burnin.max = 1e+05, MCMC.burnin.pval = 0.5, MCMC.burnin.add = 1, MCMC.prop.form = ~discord + sparse, MCMC.prop.diss = ~discord + sparse, MCMC.prop.weights.form = "default", MCMC.prop.weights.diss = "default", MCMC.prop.args.form = NULL, MCMC.prop.args.diss = NULL, MCMC.maxedges = Inf, MCMC.maxchanges = 1e+06, term.options = NULL, MCMC.packagenames = c() ) control.simulate.stergm( MCMC.burnin.min = NULL, MCMC.burnin.max = NULL, MCMC.burnin.pval = NULL, MCMC.burnin.add = NULL, MCMC.prop.form = NULL, MCMC.prop.diss = NULL, MCMC.prop.weights.form = NULL, MCMC.prop.weights.diss = NULL, MCMC.prop.args.form = NULL, MCMC.prop.args.diss = NULL, MCMC.maxedges = NULL, MCMC.maxchanges = NULL, term.options = NULL, MCMC.packagenames = NULL )control.simulate.network( MCMC.burnin.min = 1000, MCMC.burnin.max = 1e+05, MCMC.burnin.pval = 0.5, MCMC.burnin.add = 1, MCMC.prop.form = ~discord + sparse, MCMC.prop.diss = ~discord + sparse, MCMC.prop.weights.form = "default", MCMC.prop.weights.diss = "default", MCMC.prop.args.form = NULL, MCMC.prop.args.diss = NULL, MCMC.maxedges = Inf, MCMC.maxchanges = 1e+06, term.options = NULL, MCMC.packagenames = c() ) control.simulate.stergm( MCMC.burnin.min = NULL, MCMC.burnin.max = NULL, MCMC.burnin.pval = NULL, MCMC.burnin.add = NULL, MCMC.prop.form = NULL, MCMC.prop.diss = NULL, MCMC.prop.weights.form = NULL, MCMC.prop.weights.diss = NULL, MCMC.prop.args.form = NULL, MCMC.prop.args.diss = NULL, MCMC.maxedges = NULL, MCMC.maxchanges = NULL, term.options = NULL, MCMC.packagenames = NULL )
MCMC.burnin.min, MCMC.burnin.max, MCMC.burnin.pval, MCMC.burnin.add
|
Number of Metropolis-Hastings steps per time step used in simulation. By default, this
is determined adaptively by keeping track of increments in the
Hamming distance between the transitioned-from network and the
network being sampled. Once To use a fixed number of steps, set |
MCMC.prop.form |
Hints and/or constraints for selecting and initializing the proposal. |
MCMC.prop.weights.form |
Specifies the proposal weighting scheme to
be used in the MCMC Metropolis-Hastings algorithm. Possible
choices may be determined by calling |
MCMC.prop.weights.diss, MCMC.prop.args.diss, MCMC.prop.diss
|
Ignored. These are included
for backwards compatibility of calls to |
MCMC.prop.args.form |
An alternative, direct way of specifying additional arguments to proposals. |
MCMC.maxedges |
The maximum number of edges that may occur during the MCMC sampling. If this number is exceeded at any time, sampling is stopped immediately. |
MCMC.maxchanges |
Maximum number of changes for which to allocate space. |
term.options |
A list of additional arguments to be passed to term initializers. See |
MCMC.packagenames |
Names of packages in which to look for change statistic functions in addition to those autodetected. This argument should not be needed outside of very strange setups. |
This function is only used within a call to the simulate()
function. See the Usage section in simulate.stergm() for
details.
These functions are included for backwards compatibility, and users are
encouraged to use control.simulate.tergm or
control.simulate.formula.tergm with the simulate.tergm()
family of functions instead. When a
control.simulate.stergm or control.simulate.network object
is passed to one of the simulate.stergm() functions,
the corresponding simulate.tergm() function is invoked,
and uses the formation proposal control arguments, ignoring the
dissolution proposal control arguments.
Note: The old dissolution formula in stergm represents
tie persistence. As a result it maps to the new Persist() operator
in tergm, NOT the Diss() operator
A list with arguments as components.
simulate.stergm(),
simulate.formula(). control.stergm()
performs a similar function for stergm().
Auxiliary function as user interface for fine-tuning TERGM simulation.
control.simulate.tergm( MCMC.burnin.min = NULL, MCMC.burnin.max = NULL, MCMC.burnin.pval = NULL, MCMC.burnin.add = NULL, MCMC.prop = NULL, MCMC.prop.weights = NULL, MCMC.prop.args = NULL, MCMC.maxedges = NULL, MCMC.maxchanges = NULL, term.options = NULL, MCMC.packagenames = NULL ) control.simulate.formula.tergm( MCMC.burnin.min = 1000, MCMC.burnin.max = 1e+05, MCMC.burnin.pval = 0.5, MCMC.burnin.add = 1, MCMC.prop = ~discord + sparse, MCMC.prop.weights = "default", MCMC.prop.args = NULL, MCMC.maxedges = Inf, MCMC.maxchanges = 1e+06, term.options = NULL, MCMC.packagenames = c() )control.simulate.tergm( MCMC.burnin.min = NULL, MCMC.burnin.max = NULL, MCMC.burnin.pval = NULL, MCMC.burnin.add = NULL, MCMC.prop = NULL, MCMC.prop.weights = NULL, MCMC.prop.args = NULL, MCMC.maxedges = NULL, MCMC.maxchanges = NULL, term.options = NULL, MCMC.packagenames = NULL ) control.simulate.formula.tergm( MCMC.burnin.min = 1000, MCMC.burnin.max = 1e+05, MCMC.burnin.pval = 0.5, MCMC.burnin.add = 1, MCMC.prop = ~discord + sparse, MCMC.prop.weights = "default", MCMC.prop.args = NULL, MCMC.maxedges = Inf, MCMC.maxchanges = 1e+06, term.options = NULL, MCMC.packagenames = c() )
MCMC.burnin.min, MCMC.burnin.max, MCMC.burnin.pval, MCMC.burnin.add
|
Number of Metropolis-Hastings steps
per time step used in simulation. By default, this
is determined adaptively by keeping track of increments in the
Hamming distance between the transitioned-from network and the
network being sampled. Once To use a fixed number of steps, set |
MCMC.prop |
Hints and/or constraints for selecting and initializing the proposal. |
MCMC.prop.weights |
Specifies the proposal weighting scheme to
be used in the MCMC Metropolis-Hastings algorithm. Possible
choices may be determined by calling |
MCMC.prop.args |
An alternative, direct way of specifying additional arguments to the proposal. |
MCMC.maxedges |
The maximum number of edges that may occur during the MCMC sampling. If this number is exceeded at any time, sampling is stopped immediately. |
MCMC.maxchanges |
Maximum number of changes for which to allocate space. |
term.options |
A list of additional arguments to be passed to term initializers. See |
MCMC.packagenames |
Names of packages in which to look for change statistic functions in addition to those autodetected. This argument should not be needed outside of very strange setups. |
This function is only used within a call to the simulate()
function. See the Usage section in simulate.tergm() for
details.
A list with arguments as components.
simulate.tergm(),
simulate.formula(). control.tergm()
performs a similar function for tergm().
Auxiliary function as user interface for fine-tuning 'stergm' fitting.
control.stergm( init.form = NULL, init.diss = NULL, init.method = NULL, force.main = FALSE, MCMC.prop.form = ~discord + sparse, MCMC.prop.diss = ~discord + sparse, MCMC.prop.weights.form = "default", MCMC.prop.args.form = NULL, MCMC.prop.weights.diss = "default", MCMC.prop.args.diss = NULL, MCMC.maxedges = Inf, MCMC.maxchanges = 1e+06, MCMC.packagenames = c(), CMLE.MCMC.burnin = 1024 * 16, CMLE.MCMC.interval = 1024, CMLE.ergm = NULL, CMLE.form.ergm = control.ergm(init = init.form, MCMC.burnin = CMLE.MCMC.burnin, MCMC.interval = CMLE.MCMC.interval, MCMC.prop = MCMC.prop.form, MCMC.prop.weights = MCMC.prop.weights.form, MCMC.prop.args = MCMC.prop.args.form, MCMC.maxedges = MCMC.maxedges, MCMC.packagenames = MCMC.packagenames, parallel = parallel, parallel.type = parallel.type, parallel.version.check = parallel.version.check, parallel.inherit.MT = parallel.inherit.MT, force.main = force.main), CMLE.diss.ergm = control.ergm(init = init.diss, MCMC.burnin = CMLE.MCMC.burnin, MCMC.interval = CMLE.MCMC.interval, MCMC.prop = MCMC.prop.diss, MCMC.prop.weights = MCMC.prop.weights.diss, MCMC.prop.args = MCMC.prop.args.diss, MCMC.maxedges = MCMC.maxedges, MCMC.packagenames = MCMC.packagenames, parallel = parallel, parallel.type = parallel.type, parallel.version.check = parallel.version.check, parallel.inherit.MT = parallel.inherit.MT, force.main = force.main), CMLE.NA.impute = c(), CMLE.term.check.override = FALSE, EGMME.main.method = c("Gradient-Descent"), EGMME.initialfit.control = control.ergm(), EGMME.MCMC.burnin.min = 1000, EGMME.MCMC.burnin.max = 1e+05, EGMME.MCMC.burnin.pval = 0.5, EGMME.MCMC.burnin.add = 1, MCMC.burnin = NULL, MCMC.burnin.mul = NULL, SAN.maxit = 4, SAN.nsteps.times = 8, SAN = control.san(term.options = term.options, SAN.maxit = SAN.maxit, SAN.prop = MCMC.prop.form, SAN.prop.weights = MCMC.prop.weights.form, SAN.prop.args = MCMC.prop.args.form, SAN.nsteps = round(sqrt(EGMME.MCMC.burnin.min * EGMME.MCMC.burnin.max)) * SAN.nsteps.times, SAN.packagenames = MCMC.packagenames, parallel = parallel, parallel.type = parallel.type, parallel.version.check = parallel.version.check, parallel.inherit.MT = FALSE), SA.restarts = 10, SA.burnin = 1000, SA.plot.progress = FALSE, SA.max.plot.points = 400, SA.plot.stats = FALSE, SA.init.gain = 0.1, SA.gain.decay = 0.5, SA.runlength = 25, SA.interval.mul = 2, SA.init.interval = 500, SA.min.interval = 20, SA.max.interval = 500, SA.phase1.minruns = 4, SA.phase1.tries = 20, SA.phase1.jitter = 0.1, SA.phase1.max.q = 0.1, SA.phase1.backoff.rat = 1.05, SA.phase2.levels.max = 40, SA.phase2.levels.min = 4, SA.phase2.max.mc.se = 0.001, SA.phase2.repeats = 400, SA.stepdown.maxn = 200, SA.stepdown.p = 0.05, SA.stop.p = 0.1, SA.stepdown.ct = 5, SA.phase2.backoff.rat = 1.1, SA.keep.oh = 0.5, SA.keep.min.runs = 8, SA.keep.min = 0, SA.phase2.jitter.mul = 0.2, SA.phase2.maxreljump = 4, SA.guard.mul = 4, SA.par.eff.pow = 1, SA.robust = FALSE, SA.oh.memory = 1e+05, SA.refine = c("mean", "linear", "none"), SA.se = TRUE, SA.phase3.samplesize.runs = 10, SA.restart.on.err = TRUE, term.options = NULL, seed = NULL, parallel = 0, parallel.type = NULL, parallel.version.check = TRUE, parallel.inherit.MT = FALSE, ... )control.stergm( init.form = NULL, init.diss = NULL, init.method = NULL, force.main = FALSE, MCMC.prop.form = ~discord + sparse, MCMC.prop.diss = ~discord + sparse, MCMC.prop.weights.form = "default", MCMC.prop.args.form = NULL, MCMC.prop.weights.diss = "default", MCMC.prop.args.diss = NULL, MCMC.maxedges = Inf, MCMC.maxchanges = 1e+06, MCMC.packagenames = c(), CMLE.MCMC.burnin = 1024 * 16, CMLE.MCMC.interval = 1024, CMLE.ergm = NULL, CMLE.form.ergm = control.ergm(init = init.form, MCMC.burnin = CMLE.MCMC.burnin, MCMC.interval = CMLE.MCMC.interval, MCMC.prop = MCMC.prop.form, MCMC.prop.weights = MCMC.prop.weights.form, MCMC.prop.args = MCMC.prop.args.form, MCMC.maxedges = MCMC.maxedges, MCMC.packagenames = MCMC.packagenames, parallel = parallel, parallel.type = parallel.type, parallel.version.check = parallel.version.check, parallel.inherit.MT = parallel.inherit.MT, force.main = force.main), CMLE.diss.ergm = control.ergm(init = init.diss, MCMC.burnin = CMLE.MCMC.burnin, MCMC.interval = CMLE.MCMC.interval, MCMC.prop = MCMC.prop.diss, MCMC.prop.weights = MCMC.prop.weights.diss, MCMC.prop.args = MCMC.prop.args.diss, MCMC.maxedges = MCMC.maxedges, MCMC.packagenames = MCMC.packagenames, parallel = parallel, parallel.type = parallel.type, parallel.version.check = parallel.version.check, parallel.inherit.MT = parallel.inherit.MT, force.main = force.main), CMLE.NA.impute = c(), CMLE.term.check.override = FALSE, EGMME.main.method = c("Gradient-Descent"), EGMME.initialfit.control = control.ergm(), EGMME.MCMC.burnin.min = 1000, EGMME.MCMC.burnin.max = 1e+05, EGMME.MCMC.burnin.pval = 0.5, EGMME.MCMC.burnin.add = 1, MCMC.burnin = NULL, MCMC.burnin.mul = NULL, SAN.maxit = 4, SAN.nsteps.times = 8, SAN = control.san(term.options = term.options, SAN.maxit = SAN.maxit, SAN.prop = MCMC.prop.form, SAN.prop.weights = MCMC.prop.weights.form, SAN.prop.args = MCMC.prop.args.form, SAN.nsteps = round(sqrt(EGMME.MCMC.burnin.min * EGMME.MCMC.burnin.max)) * SAN.nsteps.times, SAN.packagenames = MCMC.packagenames, parallel = parallel, parallel.type = parallel.type, parallel.version.check = parallel.version.check, parallel.inherit.MT = FALSE), SA.restarts = 10, SA.burnin = 1000, SA.plot.progress = FALSE, SA.max.plot.points = 400, SA.plot.stats = FALSE, SA.init.gain = 0.1, SA.gain.decay = 0.5, SA.runlength = 25, SA.interval.mul = 2, SA.init.interval = 500, SA.min.interval = 20, SA.max.interval = 500, SA.phase1.minruns = 4, SA.phase1.tries = 20, SA.phase1.jitter = 0.1, SA.phase1.max.q = 0.1, SA.phase1.backoff.rat = 1.05, SA.phase2.levels.max = 40, SA.phase2.levels.min = 4, SA.phase2.max.mc.se = 0.001, SA.phase2.repeats = 400, SA.stepdown.maxn = 200, SA.stepdown.p = 0.05, SA.stop.p = 0.1, SA.stepdown.ct = 5, SA.phase2.backoff.rat = 1.1, SA.keep.oh = 0.5, SA.keep.min.runs = 8, SA.keep.min = 0, SA.phase2.jitter.mul = 0.2, SA.phase2.maxreljump = 4, SA.guard.mul = 4, SA.par.eff.pow = 1, SA.robust = FALSE, SA.oh.memory = 1e+05, SA.refine = c("mean", "linear", "none"), SA.se = TRUE, SA.phase3.samplesize.runs = 10, SA.restart.on.err = TRUE, term.options = NULL, seed = NULL, parallel = 0, parallel.type = NULL, parallel.version.check = TRUE, parallel.inherit.MT = FALSE, ... )
init.form, init.diss
|
numeric or
Passing coefficients from a previous run can be used to
"resume" an uncoverged |
init.method |
Estimation method used to acquire initial values
for estimation. If |
force.main |
Logical: If TRUE, then force MCMC-based estimation method, even if the exact MLE can be computed via maximum pseudolikelihood estimation. |
MCMC.prop.form |
Hints and/or constraints for selecting and initializing the proposal. |
MCMC.prop.weights.form |
Specifies the proposal weighting to use. |
MCMC.prop.args.form |
A direct way of specifying arguments to the proposal. |
MCMC.prop.weights.diss, MCMC.prop.args.diss, MCMC.prop.diss
|
Ignored. |
MCMC.maxedges |
The maximum number of edges that may occur during the MCMC sampling. If this number is exceeded at any time, sampling is stopped immediately. |
MCMC.maxchanges |
Maximum number of changes in dynamic network simulation for which to allocate space. |
MCMC.packagenames |
Names of packages in which to look for change statistic functions in addition to those autodetected. This argument should not be needed outside of very strange setups. |
CMLE.MCMC.burnin |
Burnin used in CMLE fitting. |
CMLE.MCMC.interval |
Number of Metropolis-Hastings steps between successive draws when running MCMC MLE. |
CMLE.ergm |
A convenience argument for specifying both
|
CMLE.form.ergm |
Control parameters used to fit the CMLE. See
|
CMLE.diss.ergm |
Ignored, with the exception of initial parameter values. |
CMLE.NA.impute |
In STERGM CMLE, missing dyads in
transitioned-to networks are accommodated using methods of
Handcock and Gile (2009), but a similar approach to
transitioned-from networks requires much more complex methods
that are not, currently, implemented. By default, no imputation is performed, and the fitting stops with an error if any transitioned-from networks have missing dyads. |
CMLE.term.check.override |
The method
|
EGMME.main.method |
Estimation method used to find the Equilibrium Generalized Method of Moments estimator. Currently only "Gradient-Descent" is implemented. |
EGMME.initialfit.control |
Control object for the ergm fit in tergm.EGMME.initialfit |
EGMME.MCMC.burnin.min, EGMME.MCMC.burnin.max
|
Number of
Metropolis-Hastings steps
per time step used in EGMME fitting. By default, this is
determined adaptively by keeping track of increments in the
Hamming distance between the transitioned-from network and the
network being sampled. Once To use a fixed number of steps, set
|
EGMME.MCMC.burnin.pval, EGMME.MCMC.burnin.add
|
Number of
Metropolis-Hastings steps
per time step used in EGMME fitting. By default, this is
determined adaptively by keeping track of increments in the
Hamming distance between the transitioned-from network and the
network being sampled. Once To use a fixed number of steps, set
|
MCMC.burnin, MCMC.burnin.mul
|
No longer used. See
|
SAN.maxit |
When |
SAN.nsteps.times |
Multiplier for |
SAN |
SAN control parameters. See
|
SA.restarts |
Maximum number of times to restart a failed optimization process. |
SA.burnin |
Number of time steps to advance the starting network before beginning the optimization. |
SA.plot.progress, SA.plot.stats
|
Logical: Plot information
about the fit as it proceeds. If Do NOT use these with non-interactive plotting devices like
|
SA.max.plot.points |
If |
SA.init.gain |
Initial gain, the multiplier for the parameter update size. If the process initially goes crazy beyond recovery, lower this value. |
SA.gain.decay |
Gain decay factor. |
SA.runlength |
Number of parameter trials and updates per C run. |
SA.interval.mul |
The number of time steps between updates of the parameters is set to be this times the mean duration of extant ties. |
SA.init.interval |
Initial number of time steps between updates of the parameters. |
SA.min.interval, SA.max.interval
|
Upper and lower bounds on the number of time steps between updates of the parameters. |
SA.phase1.minruns |
Number of runs during Phase 1 for estimating the gradient, before every gradient update. |
SA.phase1.tries |
Number of runs trying to find a reasonable parameter and network configuration. |
SA.phase1.jitter |
Initial jitter standard deviation of each parameter. |
SA.phase1.max.q |
Q-value (false discovery rate) that a gradient estimate must obtain before it is accepted (since sign is what is important). |
SA.phase1.backoff.rat, SA.phase2.backoff.rat
|
If the run produces this relative increase in the approximate objective function, it will be backed off. |
SA.phase2.levels.min, SA.phase2.levels.max
|
Range of gain levels (subphases) to go through. |
SA.phase2.max.mc.se |
Approximate precision of the estimates that must be attained before stopping. |
SA.phase2.repeats, SA.stepdown.maxn
|
A gain level may be
repeated multiple times (up to |
SA.stepdown.p, SA.stepdown.ct
|
A gain level may be repeated
multiple times (up to |
SA.stop.p |
At the end of each gain level after the minimum, if the precision is sufficiently high, the relationship between the parameters and the targets is tested for evidence of local nonlinearity. This is the p-value used. If that test fails to reject, a Phase 3 run is made with the new parameter values, and the estimating equations are tested for difference from 0. If this test fails to reject, the optimization is finished. If either of these tests rejects, at |
SA.keep.oh, SA.keep.min, SA.keep.min.runs
|
Parameters controlling how much of optimization history to keep for gradient and covariance estimation. A history record will be kept if it's at least one of the following:
|
SA.phase2.jitter.mul |
Jitter standard deviation of each parameter is this value times its standard deviation without jitter. |
SA.phase2.maxreljump |
To keep the optimization from "running away" due to, say, a poor gradient estimate building on itself, if a magnitude of change (Mahalanobis distance) in parameters over the course of a run divided by average magnitude of change for recent runs exceeds this, the change is truncated to this amount times the average for recent runs. |
SA.guard.mul |
The multiplier for the range of parameter and statistics values to compute the guard width. |
SA.par.eff.pow |
Because some parameters have much, much
greater effects than others, it improves numerical conditioning
and makes estimation more stable to rescale the |
SA.robust |
Whether to use robust linear regression (for gradients) and covariance estimation. |
SA.oh.memory |
Absolute maximum number of data points per thread to store in the full optimization history. |
SA.refine |
Method, if any, used to refine the point estimate at the end: "linear" for linear interpolation, "mean" for average, and "none" to use the last value. |
SA.se |
Logical: If TRUE (the default), get an MCMC sample of
statistics at the final estimate and compute the covariance
matrix (and hence standard errors) of the parameters. This sample
is stored and can also be used by
|
SA.phase3.samplesize.runs |
This many optimization runs will be used to determine whether the optimization has converged and to estimate the standard errors. |
SA.restart.on.err |
Logical: if |
term.options |
A list of additional arguments to be passed to term initializers. See |
seed |
Seed value (integer) for the random number generator. See
|
parallel |
Number of threads in which to run the
sampling. Defaults to 0 (no parallelism). See |
parallel.type |
API to use for parallel processing. Defaults
to using the parallel package with PSOCK clusters. See
|
parallel.version.check |
Logical: If TRUE, check that the version of ergm running on the slave nodes is the same as that running on the master node. |
parallel.inherit.MT |
Logical: If TRUE, slave nodes and
processes inherit the |
... |
Additional arguments, passed to other functions This argument is helpful because it collects any control parameters that have been deprecated; a warning message is printed in case of deprecated arguments. |
This function is only used within a call to the stergm()
function. See the Usage section in stergm() for details.
Generally speaking, control.stergm is remapped to control.tergm,
with dissolution controls ignored and formation controls used as controls
for the overall tergm process. An exception to this rule is the
initial parameter values specified via init.form, init.diss,
CMLE.form.ergm$init, and CMLE.diss.ergm$init, which will be
remapped jointly with the stergm() arguments offset.coef.form
and offset.coef.diss to determine the initial parameter values passed
to tergm.
It is recommended that new code make use of tergm and control.tergm
directly; stergm wrappers are included only for backwards compatibility.
A list with arguments as components.
Boer, P., Huisman, M., Snijders, T.A.B., and Zeggelink, E.P.H. (2003), StOCNET User\'s Manual. Version 1.4.
Firth (1993), Bias Reduction in Maximum Likelihood Estimates. Biometrika, 80: 27-38.
Hunter, D. R. and M. S. Handcock (2006), Inference in curved exponential family models for networks. Journal of Computational and Graphical Statistics, 15: 565-583.
Hummel, R. M., Hunter, D. R., and Handcock, M. S. (2010), A Steplength Algorithm for Fitting ERGMs, Penn State Department of Statistics Technical Report.
stergm(), tergm(), control.tergm(). The
control.simulate.stergm() function performs a similar
function for simulate.tergm().
Auxiliary function as user interface for fine-tuning 'tergm' fitting.
control.tergm( init = NULL, init.method = NULL, force.main = FALSE, MCMC.prop = ~discord + sparse, MCMC.prop.weights = "default", MCMC.prop.args = NULL, MCMC.maxedges = Inf, MCMC.maxchanges = 1e+06, MCMC.packagenames = c(), CMLE.MCMC.burnin = 1024 * 16, CMLE.MCMC.interval = 1024, CMLE.ergm = control.ergm(init = init, MCMC.burnin = CMLE.MCMC.burnin, MCMC.interval = CMLE.MCMC.interval, MCMC.prop = MCMC.prop, MCMC.prop.weights = MCMC.prop.weights, MCMC.prop.args = MCMC.prop.args, MCMC.maxedges = MCMC.maxedges, MCMC.packagenames = MCMC.packagenames, parallel = parallel, parallel.type = parallel.type, parallel.version.check = parallel.version.check, force.main = force.main, term.options = term.options), CMLE.NA.impute = c(), CMLE.term.check.override = FALSE, EGMME.main.method = c("Gradient-Descent"), EGMME.initialfit.control = control.ergm(), EGMME.MCMC.burnin.min = 1000, EGMME.MCMC.burnin.max = 1e+05, EGMME.MCMC.burnin.pval = 0.5, EGMME.MCMC.burnin.add = 1, MCMC.burnin = NULL, MCMC.burnin.mul = NULL, SAN.maxit = 4, SAN.nsteps.times = 8, SAN = control.san(term.options = term.options, SAN.maxit = SAN.maxit, SAN.prop = MCMC.prop, SAN.prop.weights = MCMC.prop.weights, SAN.prop.args = MCMC.prop.args, SAN.nsteps = round(sqrt(EGMME.MCMC.burnin.min * EGMME.MCMC.burnin.max)) * SAN.nsteps.times, SAN.packagenames = MCMC.packagenames, parallel = parallel, parallel.type = parallel.type, parallel.version.check = parallel.version.check, parallel.inherit.MT = parallel.inherit.MT), SA.restarts = 10, SA.burnin = 1000, SA.plot.progress = FALSE, SA.max.plot.points = 400, SA.plot.stats = FALSE, SA.init.gain = 0.1, SA.gain.decay = 0.5, SA.runlength = 25, SA.interval.mul = 2, SA.init.interval = 500, SA.min.interval = 20, SA.max.interval = 500, SA.phase1.minruns = 4, SA.phase1.tries = 20, SA.phase1.jitter = 0.1, SA.phase1.max.q = 0.1, SA.phase1.backoff.rat = 1.05, SA.phase2.levels.max = 40, SA.phase2.levels.min = 4, SA.phase2.max.mc.se = 0.001, SA.phase2.repeats = 400, SA.stepdown.maxn = 200, SA.stepdown.p = 0.05, SA.stop.p = 0.1, SA.stepdown.ct = 5, SA.phase2.backoff.rat = 1.1, SA.keep.oh = 0.5, SA.keep.min.runs = 8, SA.keep.min = 0, SA.phase2.jitter.mul = 0.2, SA.phase2.maxreljump = 4, SA.guard.mul = 4, SA.par.eff.pow = 1, SA.robust = FALSE, SA.oh.memory = 1e+05, SA.refine = c("mean", "linear", "none"), SA.se = TRUE, SA.phase3.samplesize.runs = 10, SA.restart.on.err = TRUE, term.options = NULL, seed = NULL, parallel = 0, parallel.type = NULL, parallel.version.check = TRUE, parallel.inherit.MT = FALSE )control.tergm( init = NULL, init.method = NULL, force.main = FALSE, MCMC.prop = ~discord + sparse, MCMC.prop.weights = "default", MCMC.prop.args = NULL, MCMC.maxedges = Inf, MCMC.maxchanges = 1e+06, MCMC.packagenames = c(), CMLE.MCMC.burnin = 1024 * 16, CMLE.MCMC.interval = 1024, CMLE.ergm = control.ergm(init = init, MCMC.burnin = CMLE.MCMC.burnin, MCMC.interval = CMLE.MCMC.interval, MCMC.prop = MCMC.prop, MCMC.prop.weights = MCMC.prop.weights, MCMC.prop.args = MCMC.prop.args, MCMC.maxedges = MCMC.maxedges, MCMC.packagenames = MCMC.packagenames, parallel = parallel, parallel.type = parallel.type, parallel.version.check = parallel.version.check, force.main = force.main, term.options = term.options), CMLE.NA.impute = c(), CMLE.term.check.override = FALSE, EGMME.main.method = c("Gradient-Descent"), EGMME.initialfit.control = control.ergm(), EGMME.MCMC.burnin.min = 1000, EGMME.MCMC.burnin.max = 1e+05, EGMME.MCMC.burnin.pval = 0.5, EGMME.MCMC.burnin.add = 1, MCMC.burnin = NULL, MCMC.burnin.mul = NULL, SAN.maxit = 4, SAN.nsteps.times = 8, SAN = control.san(term.options = term.options, SAN.maxit = SAN.maxit, SAN.prop = MCMC.prop, SAN.prop.weights = MCMC.prop.weights, SAN.prop.args = MCMC.prop.args, SAN.nsteps = round(sqrt(EGMME.MCMC.burnin.min * EGMME.MCMC.burnin.max)) * SAN.nsteps.times, SAN.packagenames = MCMC.packagenames, parallel = parallel, parallel.type = parallel.type, parallel.version.check = parallel.version.check, parallel.inherit.MT = parallel.inherit.MT), SA.restarts = 10, SA.burnin = 1000, SA.plot.progress = FALSE, SA.max.plot.points = 400, SA.plot.stats = FALSE, SA.init.gain = 0.1, SA.gain.decay = 0.5, SA.runlength = 25, SA.interval.mul = 2, SA.init.interval = 500, SA.min.interval = 20, SA.max.interval = 500, SA.phase1.minruns = 4, SA.phase1.tries = 20, SA.phase1.jitter = 0.1, SA.phase1.max.q = 0.1, SA.phase1.backoff.rat = 1.05, SA.phase2.levels.max = 40, SA.phase2.levels.min = 4, SA.phase2.max.mc.se = 0.001, SA.phase2.repeats = 400, SA.stepdown.maxn = 200, SA.stepdown.p = 0.05, SA.stop.p = 0.1, SA.stepdown.ct = 5, SA.phase2.backoff.rat = 1.1, SA.keep.oh = 0.5, SA.keep.min.runs = 8, SA.keep.min = 0, SA.phase2.jitter.mul = 0.2, SA.phase2.maxreljump = 4, SA.guard.mul = 4, SA.par.eff.pow = 1, SA.robust = FALSE, SA.oh.memory = 1e+05, SA.refine = c("mean", "linear", "none"), SA.se = TRUE, SA.phase3.samplesize.runs = 10, SA.restart.on.err = TRUE, term.options = NULL, seed = NULL, parallel = 0, parallel.type = NULL, parallel.version.check = TRUE, parallel.inherit.MT = FALSE )
init |
numeric or
Passing coefficients from a previous run can be used to
"resume" an uncoverged |
init.method |
Estimation method used to acquire initial values
for estimation. If |
force.main |
Logical: If TRUE, then force MCMC-based estimation method, even if the exact MLE can be computed via maximum pseudolikelihood estimation. |
MCMC.prop |
Hints and/or constraints for selecting and initializing the proposal. |
MCMC.prop.weights |
Specifies the proposal weighting to use. |
MCMC.prop.args |
A direct way of specifying arguments to the proposal. |
MCMC.maxedges |
The maximum number of edges that may occur during the MCMC sampling. If this number is exceeded at any time, sampling is stopped immediately. |
MCMC.maxchanges |
Maximum number of changes permitted to occur during the simulation. |
MCMC.packagenames |
Names of packages in which to look for change statistic functions in addition to those autodetected. This argument should not be needed outside of very strange setups. |
CMLE.MCMC.burnin |
Burnin used in CMLE fitting. |
CMLE.MCMC.interval |
Number of Metropolis-Hastings steps between successive draws when running MCMC MLE. |
CMLE.ergm |
Control parameters used
to fit the CMLE. See |
CMLE.NA.impute |
In TERGM CMLE, missing dyads in
transitioned-to networks are accommodated using methods of
Handcock and Gile (2009), but a similar approach to
transitioned-from networks requires much more complex methods
that are not, currently, implemented. By default, no imputation is performed, and the fitting stops with an error if any transitioned-from networks have missing dyads. |
CMLE.term.check.override |
The method
|
EGMME.main.method |
Estimation method used to find the Equilibrium Generalized Method of Moments estimator. Currently only "Gradient-Descent" is implemented. |
EGMME.initialfit.control |
Control object for the ergm fit in tergm.EGMME.initialfit |
EGMME.MCMC.burnin.min, EGMME.MCMC.burnin.max
|
Number of
Metropolis-Hastings steps
per time step used in EGMME fitting. By default, this is
determined adaptively by keeping track of increments in the
Hamming distance between the transitioned-from network and the
network being sampled.
Once To use a fixed number of steps, set
|
EGMME.MCMC.burnin.pval, EGMME.MCMC.burnin.add
|
Number of
Metropolis-Hastings steps
per time step used in EGMME fitting. By default, this is
determined adaptively by keeping track of increments in the
Hamming distance between the transitioned-from network and the
network being sampled.
Once To use a fixed number of steps, set
|
MCMC.burnin, MCMC.burnin.mul
|
No longer used. See
|
SAN.maxit |
When |
SAN.nsteps.times |
Multiplier for |
SAN |
SAN control parameters. See
|
SA.restarts |
Maximum number of times to restart a failed optimization process. |
SA.burnin |
Number of time steps to advance the starting network before beginning the optimization. |
SA.plot.progress, SA.plot.stats
|
Logical: Plot information
about the fit as it proceeds. If Do NOT use these with non-interactive plotting devices like
|
SA.max.plot.points |
If |
SA.init.gain |
Initial gain, the multiplier for the parameter update size. If the process initially goes crazy beyond recovery, lower this value. |
SA.gain.decay |
Gain decay factor. |
SA.runlength |
Number of parameter trials and updates per C run. |
SA.interval.mul |
The number of time steps between updates of the parameters is set to be this times the mean duration of extant ties. |
SA.init.interval |
Initial number of time steps between updates of the parameters. |
SA.min.interval, SA.max.interval
|
Upper and lower bounds on the number of time steps between updates of the parameters. |
SA.phase1.minruns |
Number of runs during Phase 1 for estimating the gradient, before every gradient update. |
SA.phase1.tries |
Number of runs trying to find a reasonable parameter and network configuration. |
SA.phase1.jitter |
Initial jitter standard deviation of each parameter. |
SA.phase1.max.q |
Q-value (false discovery rate) that a gradient estimate must obtain before it is accepted (since sign is what is important). |
SA.phase1.backoff.rat, SA.phase2.backoff.rat
|
If the run produces this relative increase in the approximate objective function, it will be backed off. |
SA.phase2.levels.min, SA.phase2.levels.max
|
Range of gain levels (subphases) to go through. |
SA.phase2.max.mc.se |
Approximate precision of the estimates that must be attained before stopping. |
SA.phase2.repeats, SA.stepdown.maxn
|
A gain level may be
repeated multiple times (up to |
SA.stepdown.p, SA.stepdown.ct
|
A gain level may be repeated
multiple times (up to |
SA.stop.p |
At the end of each gain level after the minimum, if the precision is sufficiently high, the relationship between the parameters and the targets is tested for evidence of local nonlinearity. This is the p-value used. If that test fails to reject, a Phase 3 run is made with the new parameter values, and the estimating equations are tested for difference from 0. If this test fails to reject, the optimization is finished. If either of these tests rejects, at |
SA.keep.oh, SA.keep.min, SA.keep.min.runs
|
Parameters controlling how much of optimization history to keep for gradient and covariance estimation. A history record will be kept if it's at least one of the following:
|
SA.phase2.jitter.mul |
Jitter standard deviation of each parameter is this value times its standard deviation without jitter. |
SA.phase2.maxreljump |
To keep the optimization from "running away" due to, say, a poor gradient estimate building on itself, if a magnitude of change (Mahalanobis distance) in parameters over the course of a run divided by average magnitude of change for recent runs exceeds this, the change is truncated to this amount times the average for recent runs. |
SA.guard.mul |
The multiplier for the range of parameter and statistics values to compute the guard width. |
SA.par.eff.pow |
Because some parameters have much, much
greater effects than others, it improves numerical conditioning
and makes estimation more stable to rescale the |
SA.robust |
Whether to use robust linear regression (for gradients) and covariance estimation. |
SA.oh.memory |
Absolute maximum number of data points per thread to store in the full optimization history. |
SA.refine |
Method, if any, used to refine the point estimate at the end: "linear" for linear interpolation, "mean" for average, and "none" to use the last value. |
SA.se |
Logical: If TRUE (the default), get an MCMC sample of
statistics at the final estimate and compute the covariance
matrix (and hence standard errors) of the parameters. This sample
is stored and can also be used by
|
SA.phase3.samplesize.runs |
This many optimization runs will be used to determine whether the optimization has converged and to estimate the standard errors. |
SA.restart.on.err |
Logical: if |
term.options |
A list of additional arguments to be passed to term initializers. See |
seed |
Seed value (integer) for the random number generator. See
|
parallel |
Number of threads in which to run the
sampling. Defaults to 0 (no parallelism). See |
parallel.type |
API to use for parallel processing. Defaults
to using the parallel package with PSOCK clusters. See
|
parallel.version.check |
Logical: If TRUE, check that the version of ergm running on the slave nodes is the same as that running on the master node. |
parallel.inherit.MT |
Logical: If TRUE, slave nodes and
processes inherit the |
This function is only used within a call to the tergm()
function. See the Usage section in tergm() for details.
A list with arguments as components.
Boer, P., Huisman, M., Snijders, T.A.B., and Zeggelink, E.P.H. (2003), StOCNET User\'s Manual. Version 1.4.
Firth (1993), Bias Reduction in Maximum Likelihood Estimates. Biometrika, 80: 27-38.
Hunter, D. R. and M. S. Handcock (2006), Inference in curved exponential family models for networks. Journal of Computational and Graphical Statistics, 15: 565-583.
Hummel, R. M., Hunter, D. R., and Handcock, M. S. (2010), A Steplength Algorithm for Fitting ERGMs, Penn State Department of Statistics Technical Report.
tergm(). The
control.simulate.tergm() function performs a similar
function for simulate.tergm().
tergm.godfather().Returns a list of its arguments.
control.tergm.godfather(term.options = NULL)control.tergm.godfather(term.options = NULL)
term.options |
A list of additional arguments to be passed to term initializers. See |
The Crossection Operator Term
# binary: Cross( # formula, # lm = ~1, # subset = TRUE, # weights = 1, # contrasts = NULL, # offset = 0, # label = NULL # )# binary: Cross( # formula, # lm = ~1, # subset = TRUE, # weights = 1, # contrasts = NULL, # offset = 0, # label = NULL # )
formula |
a one-sided |
lm, subset, weights, contrasts, offset, label
|
|
This term accepts a model formula
and produces the corresponding model for the cross-sectional
network. It is mainly useful for CMLE estimation, and has no effect (i.e.,
Cross(~TERM) == ~TERM ) for EGMME and dynamic simulation.
ergmTerm for index of model terms currently visible to the package.
None
Average age of ties incident on nodes having degree in a given range
# binary: degrange.mean.age(from, to=+Inf, byarg=NULL, emptyval=0)# binary: degrange.mean.age(from, to=+Inf, byarg=NULL, emptyval=0)
from, to
|
vectors of distinct
integers or |
byarg |
specifies a vertex attribute (see Specifying Vertex attributes and Levels ( |
emptyval |
can be used to specify the value returned if the network does not have any actors with degree in the specified range. This is, technically, an arbitrary value, but it should not have a substantial effect unless a non-negligible fraction of networks at the parameter configuration of interest has no actors with specified degree. |
This term adds one
network statistic to the model for each element of from (or to ); the th
such statistic equals the average, among all ties incident on nodes
with degree greater than or equal to
from[i] but strictly less than to[i] , of the amount of time elapsed
since the tie's formation. The optional argument
ergmTerm for index of model terms currently visible to the package.
None
Average age of ties incident on nodes having a given degree
# binary: degree.mean.age(d, byarg=NULL, emptyval=0)# binary: degree.mean.age(d, byarg=NULL, emptyval=0)
d |
a vector of distinct integers |
byarg |
specifies a vertex attribute (see Specifying Vertex attributes and Levels ( |
emptyval |
can be used to specify the value returned if the network does not have any actors with degree in the specified range. This is, technically, an arbitrary value, but it should not have a substantial effect unless a non-negligible fraction of networks at the parameter configuration of interest has no actors with specified degree. |
This term adds one
network statistic to the model for each element in d ; the th
such statistic equals the average, among all ties incident on nodes
with degree exactly d[i] , of the amount of time elapsed
since the tie's formation. The optional argument
byarg specifies a vertex attribute (see
Specifying Vertex Attributes and Levels
for details). If specified, then separate degree
statistics are calculated for nodes having each separate
value of the attribute.
ergmTerm for index of model terms currently visible to the package.
None
Propose toggling discordant dyads with greater frequency (typically about 50 percent). May be used in dynamic fitting and simulation.
# discord# discord
ergmHint for index of constraints and hints currently visible to the package.
None
The Dissolution Operator Term
# binary: Diss( # formula, # lm = ~1, # subset = TRUE, # weights = 1, # contrasts = NULL, # offset = 0, # label = NULL # )# binary: Diss( # formula, # lm = ~1, # subset = TRUE, # weights = 1, # contrasts = NULL, # offset = 0, # label = NULL # )
formula |
a one-sided |
lm, subset, weights, contrasts, offset, label
|
|
This term accepts a model formula
and produces the corresponding model for the post-dissolution
network (same as Persist() ), but with all statistics negated.
Note: This is not the equivalent of the old style dissolution model,
because the signs of the coefficients are reversed. So a larger positive
coefficient for Diss() operator means more dissolution.
ergmTerm for index of model terms currently visible to the package.
None
Sum of ages of extant ties
# binary: edge.ages# binary: edge.ages
This term adds one statistic equaling sum, over all ties present in the network, of the amount of time elapsed since formation.
Unlike mean.age , this statistic is well-defined on
an empty network. However, if used as a target, it appears to
produce highly biased dissolution parameter estimates if the goal
is to get an intended average duration.
ergmTerm for index of model terms currently visible to the package.
None
The EdgeAges Operator Term
# binary: EdgeAges(formula)# binary: EdgeAges(formula)
formula |
cross-sectional, dyad-independent model formula |
This term accepts a cross-sectional, dyad-independent model formula. The statistics of the EdgeAges term are equal to the sum over all extant ties of the tie age times the on-toggle change statistics for the tie under the given model formula.
ergmTerm for index of model terms currently visible to the package.
None
Weighted sum of ages of extant ties
# binary: edgecov.ages(x, attrname=NULL)# binary: edgecov.ages(x, attrname=NULL)
x, attrname
|
a specification for the dyadic covariate: either one of the following, or the name of a network attribute containing one of the following:
|
This term adds one statistic equaling sum, over all ties present in the network, of the amount of time elapsed since formation, multiplied by a dyadic covariate.
"Weights" can be negative.
Unlike edgecov.mean.age , this statistic is well-defined on
an empty network. However, if used as a target, it appears to
produce highly biased dissolution parameter estimates if the goal
is to get an intended average duration.
ergmTerm for index of model terms currently visible to the package.
None
Weighted average age of an extant tie
# binary: edgecov.mean.age(x, attrname=NULL, emptyval=0)# binary: edgecov.mean.age(x, attrname=NULL, emptyval=0)
x, attrname
|
a specification for the dyadic covariate: either one of the following, or the name of a network attribute containing one of the following:
|
emptyval |
can be used to specify the value returned if the network is empty (or all extant edges have been weighted 0). This is, technically, an arbitrary value, but it should not have a substantial effect unless a non-negligible fraction of networks at the parameter configuration of interest is empty and/or if only a few dyads have nonzero weights. |
This term adds one statistic equaling the average, over all ties present in the network, of the amount of time elapsed since formation, weighted by a (nonnegative) dyadic covariate.
The behavior when there are negative weights is undefined.
ergmTerm for index of model terms currently visible to the package.
None
Number of edges with age falling into a specified range
# binary: edges.ageinterval(from, to=+Inf)# binary: edges.ageinterval(from, to=+Inf)
from, to
|
parameters to specify the lower bound and strict upper bounds. Can be scalars, vectors of the same length, or one of them must have length one, in which case it is recycled. |
This term counts the number of edges in the network for
which the time elapsed since formation is greater than or equal to
from but strictly less than to . In other words, it
is in the semiopen interval [from, to) .
ergmTerm for index of model terms currently visible to the package.
None
The Formation Operator Term
# binary: Form( # formula, # lm = ~1, # subset = TRUE, # weights = 1, # contrasts = NULL, # offset = 0, # label = NULL # )# binary: Form( # formula, # lm = ~1, # subset = TRUE, # weights = 1, # contrasts = NULL, # offset = 0, # label = NULL # )
formula |
a one-sided |
lm, subset, weights, contrasts, offset, label
|
|
This term accepts a model formula
and produces the corresponding model for the post-formation network:
effectively a network containing both previous time step's ties and ties just formed,
the union of the previous and current network. This is the equivalent of the
old-style formation model.
ergmTerm for index of model terms currently visible to the package.
None
This function takes a list of networks with missing dyads and returns a list of networks with missing dyads imputed according to a list of imputation directives.
impute.network.list( nwl, imputers = c(), nwl.prepend = list(), nwl.append = list() )impute.network.list( nwl, imputers = c(), nwl.prepend = list(), nwl.append = list() )
nwl |
A list of |
imputers |
A character vector giving one or more methods to impute missing dyads. Currenly implemented methods are as follows:
If
|
nwl.prepend |
An optional list of networks to treat as
preceding those in |
nwl.append |
An optional list of networks to treat as
following those in |
A list of networks with missing dyads imputed.
These functions test whether an ERGM is duration dependent or not.
The method for NULL always returns FALSE by
convention.
is.durational(object, ...) ## S3 method for class ''NULL'' is.durational(object, ...) ## S3 method for class 'ergm_model' is.durational(object, ...) ## S3 method for class 'ergm_state' is.durational(object, ...) ## S3 method for class 'formula' is.durational(object, response = NULL, basis = ergm.getnetwork(object), ...)is.durational(object, ...) ## S3 method for class ''NULL'' is.durational(object, ...) ## S3 method for class 'ergm_model' is.durational(object, ...) ## S3 method for class 'ergm_state' is.durational(object, ...) ## S3 method for class 'formula' is.durational(object, response = NULL, basis = ergm.getnetwork(object), ...)
object |
An ERGM formula, |
... |
Unused at this time. |
response, basis
|
See |
TRUE if the ERGM terms in the model are duration dependent;
FALSE otherwise.
is.durational(ergm_model): Test if the ergm_model has duration-dependent terms, which call for lasttoggle data structures.
is.durational(ergm_state): Test if the ergm_state has duration-dependent terms, which call for lasttoggle data structures.
A data structure used by tergm for tracking of limited information
about dyad edge histories.
The tergm package handles durational information attached to
network objects by way of the time and
lasttoggle network attributes. The lasttoggle data
structure is a 3-column matrix; the first two columns are tails
and heads (respectively) of dyads, and the third column is the last
time at which the dyad was toggled. The default last toggle time
is -INT_MAX/2. Last toggle times for non-edges are
periodically cleared in the C code. The time network
attribute is simply an integer, and together with the
lasttoggle data it determines the age of an extant
tie as time + 1 minus the last toggle time for that dyad.
The default value for time is 0.
Average age of an extant tie
# binary: mean.age(emptyval=0, log=FALSE)# binary: mean.age(emptyval=0, log=FALSE)
emptyval |
can be used to specify the value returned if the network is empty. This is, technically, an arbitrary value, but it should not have a substantial effect unless a non-negligible fraction of networks at the parameter configuration of interest is empty. |
log |
logical specifying if mean log age should be returned instead of mean age |
This term adds one statistic equaling the average, over all ties present in the network, of the amount of time elapsed since formation.
ergmTerm for index of model terms currently visible to the package.
None
A function for specifying the LHS of a temporal network series ERGM.
NetSeries(..., order = 1, NA.impute = NULL)NetSeries(..., order = 1, NA.impute = NULL)
... |
series specification, in one of three formats:
|
order |
how many previous networks to store as an accessible covariate of the model. |
NA.impute |
How missing dyads in transitioned-from networks
are be imputed when using conditional estimation. See argument
|
A network object with temporal metadata.
It is not recommended to modify the network returned by
NetSeries except by adding and removing edges, and even that
must be done with some care, to avoid putting it into an
inconsistent state.
It is almost always better to modify the original networks and regenerate the series.
ergmTerm for specific terms.
data(samplk) # Method 1: list of networks monks <- NetSeries(list(samplk1,samplk2,samplk3)) ergm(monks ~ Form(~edges)+Diss(~edges)) ergm(monks ~ Form(~edges)+Persist(~edges)) # Method 2: networks as arguments monks <- NetSeries(samplk1,samplk2,samplk3) ergm(monks ~ Form(~edges)+Diss(~edges)) ergm(monks ~ Form(~edges)+Persist(~edges)) # Method 3: networkDynamic and time points: ## TODOdata(samplk) # Method 1: list of networks monks <- NetSeries(list(samplk1,samplk2,samplk3)) ergm(monks ~ Form(~edges)+Diss(~edges)) ergm(monks ~ Form(~edges)+Persist(~edges)) # Method 2: networks as arguments monks <- NetSeries(samplk1,samplk2,samplk3) ergm(monks ~ Form(~edges)+Diss(~edges)) ergm(monks ~ Form(~edges)+Persist(~edges)) # Method 3: networkDynamic and time points: ## TODO
Average ages of extant half-ties incident on nodes of specified attribute levels
# binary: nodefactor.mean.age(attr, levels=NULL, emptyval=0, log=FALSE)# binary: nodefactor.mean.age(attr, levels=NULL, emptyval=0, log=FALSE)
attr |
a vertex attribute specification (see Specifying Vertex attributes and Levels ( |
levels |
controls what levels are included. Note that the default
|
emptyval |
can be used to specify the value returned if the network is empty. A different value may be
specified for each level of |
log |
logical specifying if mean log age should be returned instead of mean age |
This term adds one statistic for each level of attr ,
equaling the average, over all half-ties incident on nodes of that level,
of the amount of time elapsed since formation.
ergmTerm for index of model terms currently visible to the package.
None
Average ages of extant ties of specified mixing types
# binary: nodemix.mean.age(attr, b1levels=NULL, b2levels=NULL, levels=NULL, # levels2=NULL, emptyval=0, log=FALSE)# binary: nodemix.mean.age(attr, b1levels=NULL, b2levels=NULL, levels=NULL, # levels2=NULL, emptyval=0, log=FALSE)
attr |
a vertex attribute specification (see Specifying Vertex attributes and Levels ( |
b1levels, b2levels, levels, level2
|
control what statistics are included in the model and the order in which they appear. |
emptyval |
can be used to specify the value returned if the network is empty. A different value may be
specified for each mixing type of |
log |
logical specifying if mean log age should be returned instead of mean age |
This term adds one statistic for each mixing type of attr ,
equaling the average, over all ties of that mixing type,
of the amount of time elapsed since formation.
ergmTerm for index of model terms currently visible to the package.
None
The Persistence Operator Term
# binary: Persist( # formula, # lm = ~1, # subset = TRUE, # weights = 1, # contrasts = NULL, # offset = 0, # label = NULL # )# binary: Persist( # formula, # lm = ~1, # subset = TRUE, # weights = 1, # contrasts = NULL, # offset = 0, # label = NULL # )
formula |
a one-sided |
lm, subset, weights, contrasts, offset, label
|
|
This term accepts a model formula and produces the corresponding model for the post-dissolution/persistence network: effectively the network containing ties that persisted since the last time step.
This is the equivalent of the old-style dissolution model. So
a larger positive coefficient for Persist() operator means
less dissolution. It
produces the same results as the new Diss() operator, except the
signs of the coefficients are negated.
ergmTerm for index of model terms currently visible to the package.
None
The simulate.network and simulate.networkDynamic wrappers
are provided for backwards compatibility. It is recommended that new
code make use of the simulate_formula.network and
simulate_formula.networkDynamic functions instead. See
simulate.tergm() for details on these new functions.
## S3 method for class 'network' simulate( object, nsim = 1, seed = NULL, formation, dissolution, coef.form, coef.diss, constraints = ~., monitor = NULL, time.slices = 1, time.start = NULL, time.burnin = 0, time.interval = 1, time.offset = 1, control = control.simulate.network(), output = c("networkDynamic", "stats", "changes", "final", "ergm_state"), stats.form = FALSE, stats.diss = FALSE, verbose = FALSE, ... ) ## S3 method for class 'networkDynamic' simulate( object, nsim = 1, seed = NULL, formation, dissolution, coef.form = attr(object, "coef.form"), coef.diss = attr(object, "coef.diss"), constraints = ~., monitor = NULL, time.slices = 1, time.start = NULL, time.burnin = 0, time.interval = 1, time.offset = 1, control = control.simulate.network(), output = c("networkDynamic", "stats", "changes", "final", "ergm_state"), stats.form = FALSE, stats.diss = FALSE, verbose = FALSE, ... )## S3 method for class 'network' simulate( object, nsim = 1, seed = NULL, formation, dissolution, coef.form, coef.diss, constraints = ~., monitor = NULL, time.slices = 1, time.start = NULL, time.burnin = 0, time.interval = 1, time.offset = 1, control = control.simulate.network(), output = c("networkDynamic", "stats", "changes", "final", "ergm_state"), stats.form = FALSE, stats.diss = FALSE, verbose = FALSE, ... ) ## S3 method for class 'networkDynamic' simulate( object, nsim = 1, seed = NULL, formation, dissolution, coef.form = attr(object, "coef.form"), coef.diss = attr(object, "coef.diss"), constraints = ~., monitor = NULL, time.slices = 1, time.start = NULL, time.burnin = 0, time.interval = 1, time.offset = 1, control = control.simulate.network(), output = c("networkDynamic", "stats", "changes", "final", "ergm_state"), stats.form = FALSE, stats.diss = FALSE, verbose = FALSE, ... )
object |
an object of type |
nsim |
Number of replications (separate chains of networks) of the
process to run and return. The |
seed |
Seed value (integer) for the random number generator. See
|
formation, dissolution
|
One-sided |
coef.form |
Parameters for the formation model. |
coef.diss |
Parameters for the dissolution (persistence) model. |
constraints |
A formula specifying one or more constraints
on the support of the distribution of the networks being modeled. Multiple constraints
may be given, separated by “+” and “-” operators. See
The default is to have no constraints except those provided through
the Together with the model terms in the formula and the reference measure, the constraints define the distribution of networks being modeled. It is also possible to specify a proposal function directly either
by passing a string with the function's name (in which case,
arguments to the proposal should be specified through the
Note that not all possible combinations of constraints and reference measures are supported. However, for relatively simple constraints (i.e., those that simply permit or forbid specific dyads or sets of dyads from changing), arbitrary combinations should be possible. |
monitor |
A one-sided formula specifying one or more terms whose
value is to be monitored. If |
time.slices |
Number of time slices (or statistics) to return from each
replication of the dynamic process. See below for return types. Defaults to
1, which, if |
time.start |
An optional argument specifying the time point at which the simulation is to start. See Details for further information. |
time.burnin |
Number of time steps to discard before starting to collect network statistics. |
time.interval |
Number of time steps between successive recordings of network statistics. |
time.offset |
Argument specifying the offset between the point when the
state of the network is sampled ( |
control |
A list of control parameters for algorithm tuning,
constructed using
|
output |
A character vector specifying output type: one of
|
stats.form, stats.diss
|
Logical: Whether to return
formation/dissolution model statistics. This is not the recommended method:
use the |
verbose |
A logical or an integer to control the amount of
progress and diagnostic information to be printed. |
... |
Further arguments passed to or used by methods. |
Note that return values may be structured differently than in past versions.
Remember that in stergm, the dissolution formula is parameterized in
terms of tie persistence: negative coefficients imply lower rates of persistence
and postive coefficients imply higher rates. The dissolution effects are simply the
negation of these coefficients.
Because the old dissolution formula in stergm represents
tie persistence, it maps to the new Persist() operator
in the tergm function, NOT the Diss() operator
Depends on the output argument. See simulate.tergm()
for details. Note that some formation/dissolution separated
information is also attached to the return value for calls made through
simulate.network and simulate.networkDynamic in
an attempt to increase backwards compatibility.
logit<-function(p)log(p/(1-p)) coef.form.f<-function(coef.diss,density) -log(((1+exp(coef.diss))/(density/(1-density)))-1) # Construct a network with 20 nodes and 20 edges n<-20 target.stats<-edges<-20 g0<-network.initialize(n,dir=TRUE) g1<-san(g0~edges,target.stats=target.stats,verbose=TRUE) S<-10 # To get an average duration of 10... duration<-10 coef.diss<-logit(1-1/duration) # To get an average of 20 edges... dyads<-network.dyadcount(g1) density<-edges/dyads coef.form<-coef.form.f(coef.diss,density) # ... coefficients. print(coef.form) print(coef.diss) # Simulate a networkDynamic dynsim<-simulate(g1,formation=~edges,dissolution=~edges, coef.form=coef.form,coef.diss=coef.diss, time.slices=S,verbose=TRUE) # "Resume" the simulation. dynsim2<-simulate(dynsim,formation=~edges,dissolution=~edges,time.slices=S,verbose=TRUE)logit<-function(p)log(p/(1-p)) coef.form.f<-function(coef.diss,density) -log(((1+exp(coef.diss))/(density/(1-density)))-1) # Construct a network with 20 nodes and 20 edges n<-20 target.stats<-edges<-20 g0<-network.initialize(n,dir=TRUE) g1<-san(g0~edges,target.stats=target.stats,verbose=TRUE) S<-10 # To get an average duration of 10... duration<-10 coef.diss<-logit(1-1/duration) # To get an average of 20 edges... dyads<-network.dyadcount(g1) density<-edges/dyads coef.form<-coef.form.f(coef.diss,density) # ... coefficients. print(coef.form) print(coef.diss) # Simulate a networkDynamic dynsim<-simulate(g1,formation=~edges,dissolution=~edges, coef.form=coef.form,coef.diss=coef.diss, time.slices=S,verbose=TRUE) # "Resume" the simulation. dynsim2<-simulate(dynsim,formation=~edges,dissolution=~edges,time.slices=S,verbose=TRUE)
simulate() is used to draw from temporal
exponential family random network models in their natural parameterizations.
See tergm() for more information on these models.
## S3 method for class 'tergm' simulate( object, nsim = 1, seed = NULL, coef = coefficients(object), constraints = object$constraints, monitor = object$targets, time.slices = 1, time.start = NULL, time.burnin = 0, time.interval = 1, control = control.simulate.tergm(), output = c("networkDynamic", "stats", "changes", "final", "ergm_state"), nw.start = NULL, stats = FALSE, verbose = FALSE, ... ) ## S3 method for class 'network' simulate_formula( object, nsim = 1, seed = NULL, coef = NULL, constraints = ~., monitor = NULL, time.slices = 1, time.start = NULL, time.burnin = 0, time.interval = 1, time.offset = 1, control = control.simulate.formula.tergm(), output = c("networkDynamic", "stats", "changes", "final", "ergm_state"), stats = FALSE, verbose = FALSE, ..., basis = ergm.getnetwork(object), dynamic = FALSE ) ## S3 method for class 'networkDynamic' simulate_formula( object, nsim = 1, seed = NULL, coef = attr(basis, "coef"), constraints = ~., monitor = NULL, time.slices = 1, time.start = NULL, time.burnin = 0, time.interval = 1, time.offset = 1, control = control.simulate.formula.tergm(), output = c("networkDynamic", "stats", "changes", "final", "ergm_state"), stats = FALSE, verbose = FALSE, ..., basis = eval_lhs.formula(object), dynamic = FALSE )## S3 method for class 'tergm' simulate( object, nsim = 1, seed = NULL, coef = coefficients(object), constraints = object$constraints, monitor = object$targets, time.slices = 1, time.start = NULL, time.burnin = 0, time.interval = 1, control = control.simulate.tergm(), output = c("networkDynamic", "stats", "changes", "final", "ergm_state"), nw.start = NULL, stats = FALSE, verbose = FALSE, ... ) ## S3 method for class 'network' simulate_formula( object, nsim = 1, seed = NULL, coef = NULL, constraints = ~., monitor = NULL, time.slices = 1, time.start = NULL, time.burnin = 0, time.interval = 1, time.offset = 1, control = control.simulate.formula.tergm(), output = c("networkDynamic", "stats", "changes", "final", "ergm_state"), stats = FALSE, verbose = FALSE, ..., basis = ergm.getnetwork(object), dynamic = FALSE ) ## S3 method for class 'networkDynamic' simulate_formula( object, nsim = 1, seed = NULL, coef = attr(basis, "coef"), constraints = ~., monitor = NULL, time.slices = 1, time.start = NULL, time.burnin = 0, time.interval = 1, time.offset = 1, control = control.simulate.formula.tergm(), output = c("networkDynamic", "stats", "changes", "final", "ergm_state"), stats = FALSE, verbose = FALSE, ..., basis = eval_lhs.formula(object), dynamic = FALSE )
object |
for
|
nsim |
Number of replications (separate chains of networks) of the
process to run and return. The |
seed |
Seed value (integer) for the random number generator. See
|
coef |
Parameters for the model. |
constraints |
A formula specifying one or more constraints
on the support of the distribution of the networks being modeled. Multiple constraints
may be given, separated by “+” and “-” operators. See
The default is to have no constraints except those provided through
the Together with the model terms in the formula and the reference measure, the constraints define the distribution of networks being modeled. It is also possible to specify a proposal function directly either
by passing a string with the function's name (in which case,
arguments to the proposal should be specified through the
Note that not all possible combinations of constraints and reference measures are supported. However, for relatively simple constraints (i.e., those that simply permit or forbid specific dyads or sets of dyads from changing), arbitrary combinations should be possible. |
monitor |
A one-sided formula specifying one or more terms whose
value is to be monitored. If |
time.slices |
Number of time slices (or statistics) to return from each
replication of the dynamic process. See below for return types. Defaults to
1, which, if |
time.start |
An optional argument specifying the time point at which the simulation is to start. See Details for further information. |
time.burnin |
Number of time steps to discard before starting to collect network statistics. |
time.interval |
Number of time steps between successive recordings of network statistics. |
control |
A list of control parameters for algorithm tuning.
Constructed using
|
output |
A character vector specifying output type: one of
|
nw.start |
A specification for the starting network to be used by
|
stats |
Logical: Whether to return
model statistics. This is not the recommended method:
use |
verbose |
A logical or an integer to control the amount of
progress and diagnostic information to be printed. |
... |
Further arguments passed to or used by methods. |
time.offset |
Argument specifying the offset between the point when the
state of the network is sampled ( |
basis |
For the |
dynamic |
Logical; if |
The dynamic process is run forward and the results are returned. For the
method for networkDynamic, the simulation is resumed from the
last generated time point of basis (or the left hand side of object
if basis is missing), by default with the same model
and parameters.
The starting network for the tergm object method
(simulate.tergm) is determined by the nw.start argument.
If time.start is specified, it is used as the initial
time index of the simulation.
If time.start is not specified (is NULL), then
if the object carries a time stamp from which to start
or resume the simulation, either in the form
of a "time" network attribute (for the
network method — see the
lasttoggle "API") or
in the form of an net.obs.period network attribute (for the
networkDynamic method), this attribute will be used. (If
specified, time.start will override it with a warning.)
Othewise, the simulation starts at 0.
Depends on the output argument:
"stats" |
If |
"networkDynamic" |
A
When |
"changes" |
An integer matrix with four columns ( |
"final" |
A |
"ergm_state" |
The |
Note that when using simulate_formula.networkDynamic with either
"final" or "ergm_state" for output, the nodes
included in these objects are those produced by network.collapse
at the start time.
data(samplk) # Fit a transition from Time 1 to Time 2 samplk12 <- tergm(list(samplk1, samplk2)~ Form(~edges+mutual+transitiveties+cyclicalties)+ Diss(~edges+mutual+transitiveties+cyclicalties), estimate="CMLE") # direct simulation from tergm object sim1 <- simulate(samplk12, nw.start="last") # equivalent simulation from formula with network LHS; # must pass dynamic=TRUE for tergm simulation sim2 <- simulate(samplk2 ~ Form(~edges+mutual+transitiveties+cyclicalties) + Diss(~edges+mutual+transitiveties+cyclicalties), coef = coef(samplk12), dynamic=TRUE) # the default simulate output is a networkDynamic, and we can simulate # with a networkDynamic LHS as well sim3 <- simulate(sim2 ~ Form(~edges+mutual+transitiveties+cyclicalties) + Diss(~edges+mutual+transitiveties+cyclicalties), coef = coef(samplk12), dynamic=TRUE)data(samplk) # Fit a transition from Time 1 to Time 2 samplk12 <- tergm(list(samplk1, samplk2)~ Form(~edges+mutual+transitiveties+cyclicalties)+ Diss(~edges+mutual+transitiveties+cyclicalties), estimate="CMLE") # direct simulation from tergm object sim1 <- simulate(samplk12, nw.start="last") # equivalent simulation from formula with network LHS; # must pass dynamic=TRUE for tergm simulation sim2 <- simulate(samplk2 ~ Form(~edges+mutual+transitiveties+cyclicalties) + Diss(~edges+mutual+transitiveties+cyclicalties), coef = coef(samplk12), dynamic=TRUE) # the default simulate output is a networkDynamic, and we can simulate # with a networkDynamic LHS as well sim3 <- simulate(sim2 ~ Form(~edges+mutual+transitiveties+cyclicalties) + Diss(~edges+mutual+transitiveties+cyclicalties), coef = coef(samplk12), dynamic=TRUE)
A utility to facilitate argument completion of control lists, reexported from statnet.common.
This list is updated as packages are loaded and unloaded.
control.ergmdrop, init, init.method, main.method, force.main, main.hessian,
checkpoint, resume, MPLE.samplesize, init.MPLE.samplesize,
MPLE.type, MPLE.maxit, MPLE.nonvar, MPLE.nonident,
MPLE.nonident.tol, MPLE.covariance.samplesize,
MPLE.covariance.method, MPLE.covariance.sim.burnin,
MPLE.covariance.sim.interval, MPLE.check,
MPLE.constraints.ignore, MCMC.prop, MCMC.prop.weights,
MCMC.prop.args, MCMC.interval, MCMC.burnin,
MCMC.samplesize, MCMC.effectiveSize,
MCMC.effectiveSize.damp, MCMC.effectiveSize.maxruns,
MCMC.effectiveSize.burnin.pval,
MCMC.effectiveSize.burnin.min,
MCMC.effectiveSize.burnin.max,
MCMC.effectiveSize.burnin.nmin,
MCMC.effectiveSize.burnin.nmax,
MCMC.effectiveSize.burnin.PC,
MCMC.effectiveSize.burnin.scl,
MCMC.effectiveSize.order.max, MCMC.return.stats,
MCMC.runtime.traceplot, MCMC.maxedges, MCMC.addto.se,
MCMC.packagenames, SAN.maxit, SAN.nsteps.times, SAN,
MCMLE.termination, MCMLE.maxit, MCMLE.conv.min.pval,
MCMLE.confidence, MCMLE.confidence.boost,
MCMLE.confidence.boost.threshold,
MCMLE.confidence.boost.lag, MCMLE.NR.maxit,
MCMLE.NR.reltol, obs.MCMC.mul, obs.MCMC.samplesize.mul,
obs.MCMC.samplesize, obs.MCMC.effectiveSize,
obs.MCMC.interval.mul, obs.MCMC.interval,
obs.MCMC.burnin.mul, obs.MCMC.burnin, obs.MCMC.prop,
obs.MCMC.prop.weights, obs.MCMC.prop.args,
obs.MCMC.impute.min_informative,
obs.MCMC.impute.default_density, MCMLE.min.depfac,
MCMLE.sampsize.boost.pow, MCMLE.MCMC.precision,
MCMLE.MCMC.max.ESS.frac, MCMLE.metric, MCMLE.method,
MCMLE.dampening, MCMLE.dampening.min.ess,
MCMLE.dampening.level, MCMLE.steplength.margin,
MCMLE.steplength, MCMLE.steplength.parallel,
MCMLE.sequential, MCMLE.density.guard.min,
MCMLE.density.guard, MCMLE.effectiveSize,
obs.MCMLE.effectiveSize, MCMLE.interval, MCMLE.burnin,
MCMLE.samplesize.per_theta, MCMLE.samplesize.min,
MCMLE.samplesize, obs.MCMLE.samplesize.per_theta,
obs.MCMLE.samplesize.min, obs.MCMLE.samplesize,
obs.MCMLE.interval, obs.MCMLE.burnin,
MCMLE.steplength.solver, MCMLE.last.boost,
MCMLE.steplength.esteq, MCMLE.steplength.miss.sample,
MCMLE.steplength.min, MCMLE.effectiveSize.interval_drop,
MCMLE.save_intermediates, MCMLE.nonvar, MCMLE.nonident,
MCMLE.nonident.tol, SA.phase1_n, SA.initial_gain,
SA.nsubphases, SA.min_iterations, SA.max_iterations,
SA.phase3_n, SA.interval, SA.burnin, SA.samplesize,
CD.samplesize.per_theta, obs.CD.samplesize.per_theta,
CD.nsteps, CD.multiplicity, CD.nsteps.obs,
CD.multiplicity.obs, CD.maxit, CD.conv.min.pval,
CD.NR.maxit, CD.NR.reltol, CD.metric, CD.method,
CD.dampening, CD.dampening.min.ess, CD.dampening.level,
CD.steplength.margin, CD.steplength, CD.adaptive.epsilon,
CD.steplength.esteq, CD.steplength.miss.sample,
CD.steplength.min, CD.steplength.parallel,
CD.steplength.solver, loglik, term.options, seed,
parallel, parallel.type, parallel.version.check,
parallel.inherit.MT, ...
control.ergm.bridgebridge.nsteps, bridge.target.se, bridge.bidirectional, drop,
MCMC.burnin, MCMC.burnin.between, MCMC.interval,
MCMC.samplesize, obs.MCMC.burnin,
obs.MCMC.burnin.between, obs.MCMC.interval,
obs.MCMC.samplesize, MCMC.prop, MCMC.prop.weights,
MCMC.prop.args, obs.MCMC.prop,
obs.MCMC.prop.weights, obs.MCMC.prop.args,
MCMC.maxedges, MCMC.packagenames, term.options,
seed, parallel, parallel.type,
parallel.version.check, parallel.inherit.MT, ...
control.ergm.godfatherterm.options
control.gof.ergmnsim, MCMC.burnin, MCMC.interval, MCMC.batch, MCMC.prop,
MCMC.prop.weights, MCMC.prop.args, MCMC.maxedges,
MCMC.packagenames, MCMC.runtime.traceplot,
network.output, seed, parallel, parallel.type,
parallel.version.check, parallel.inherit.MT
control.gof.formulansim, MCMC.burnin, MCMC.interval, MCMC.batch, MCMC.prop,
MCMC.prop.weights, MCMC.prop.args, MCMC.maxedges,
MCMC.packagenames, MCMC.runtime.traceplot,
network.output, seed, parallel, parallel.type,
parallel.version.check, parallel.inherit.MT
control.logLik.ergmbridge.nsteps, bridge.target.se, bridge.bidirectional, drop,
MCMC.burnin, MCMC.interval, MCMC.samplesize,
obs.MCMC.samplesize, obs.MCMC.interval,
obs.MCMC.burnin, MCMC.prop, MCMC.prop.weights,
MCMC.prop.args, obs.MCMC.prop,
obs.MCMC.prop.weights, obs.MCMC.prop.args,
MCMC.maxedges, MCMC.packagenames, term.options,
seed, parallel, parallel.type,
parallel.version.check, parallel.inherit.MT, ...
control.sanSAN.maxit, SAN.tau, SAN.invcov, SAN.invcov.diag, SAN.nsteps.alloc,
SAN.nsteps, SAN.samplesize, SAN.prop, SAN.prop.weights,
SAN.prop.args, SAN.packagenames, SAN.ignore.finite.offsets,
term.options, seed, parallel, parallel.type,
parallel.version.check, parallel.inherit.MT
control.simulateMCMC.burnin, MCMC.interval, MCMC.prop, MCMC.prop.weights,
MCMC.prop.args, MCMC.batch, MCMC.effectiveSize,
MCMC.effectiveSize.damp, MCMC.effectiveSize.maxruns,
MCMC.effectiveSize.burnin.pval,
MCMC.effectiveSize.burnin.min,
MCMC.effectiveSize.burnin.max,
MCMC.effectiveSize.burnin.nmin,
MCMC.effectiveSize.burnin.nmax,
MCMC.effectiveSize.burnin.PC,
MCMC.effectiveSize.burnin.scl,
MCMC.effectiveSize.order.max, MCMC.maxedges,
MCMC.packagenames, MCMC.runtime.traceplot,
network.output, term.options, parallel, parallel.type,
parallel.version.check, parallel.inherit.MT, ...
control.simulate.ergmMCMC.burnin, MCMC.interval, MCMC.scale, MCMC.prop, MCMC.prop.weights,
MCMC.prop.args, MCMC.batch, MCMC.effectiveSize,
MCMC.effectiveSize.damp,
MCMC.effectiveSize.maxruns,
MCMC.effectiveSize.burnin.pval,
MCMC.effectiveSize.burnin.min,
MCMC.effectiveSize.burnin.max,
MCMC.effectiveSize.burnin.nmin,
MCMC.effectiveSize.burnin.nmax,
MCMC.effectiveSize.burnin.PC,
MCMC.effectiveSize.burnin.scl,
MCMC.effectiveSize.order.max, MCMC.maxedges,
MCMC.packagenames, MCMC.runtime.traceplot,
network.output, term.options, parallel,
parallel.type, parallel.version.check,
parallel.inherit.MT, ...
control.simulate.formulaMCMC.burnin, MCMC.interval, MCMC.prop, MCMC.prop.weights,
MCMC.prop.args, MCMC.batch,
MCMC.effectiveSize, MCMC.effectiveSize.damp,
MCMC.effectiveSize.maxruns,
MCMC.effectiveSize.burnin.pval,
MCMC.effectiveSize.burnin.min,
MCMC.effectiveSize.burnin.max,
MCMC.effectiveSize.burnin.nmin,
MCMC.effectiveSize.burnin.nmax,
MCMC.effectiveSize.burnin.PC,
MCMC.effectiveSize.burnin.scl,
MCMC.effectiveSize.order.max, MCMC.maxedges,
MCMC.packagenames, MCMC.runtime.traceplot,
network.output, term.options, parallel,
parallel.type, parallel.version.check,
parallel.inherit.MT, ...
control.simulate.formula.ergmMCMC.burnin, MCMC.interval, MCMC.prop, MCMC.prop.weights,
MCMC.prop.args, MCMC.batch,
MCMC.effectiveSize,
MCMC.effectiveSize.damp,
MCMC.effectiveSize.maxruns,
MCMC.effectiveSize.burnin.pval,
MCMC.effectiveSize.burnin.min,
MCMC.effectiveSize.burnin.max,
MCMC.effectiveSize.burnin.nmin,
MCMC.effectiveSize.burnin.nmax,
MCMC.effectiveSize.burnin.PC,
MCMC.effectiveSize.burnin.scl,
MCMC.effectiveSize.order.max,
MCMC.maxedges, MCMC.packagenames,
MCMC.runtime.traceplot, network.output,
term.options, parallel, parallel.type,
parallel.version.check,
parallel.inherit.MT, ...
stergm() fits Separable Temporal ERGMs'
(STERGMs) Conditional MLE (CMLE) (Krivitsky and Handcock, 2014) and
Equilibrium Generalized Method of Moments Estimator (EGMME)
(Krivitsky, 2009). This function is deprecated in favor of
tergm(), whose special case it is, and may be removed in a future
version.
stergm( nw, formation, dissolution, constraints = ~., estimate, times = NULL, offset.coef.form = NULL, offset.coef.diss = NULL, targets = NULL, target.stats = NULL, eval.loglik = NVL(getOption("tergm.eval.loglik"), getOption("ergm.eval.loglik")), control = control.stergm(), verbose = FALSE, ..., SAN.offsets = NULL )stergm( nw, formation, dissolution, constraints = ~., estimate, times = NULL, offset.coef.form = NULL, offset.coef.diss = NULL, targets = NULL, target.stats = NULL, eval.loglik = NVL(getOption("tergm.eval.loglik"), getOption("ergm.eval.loglik")), control = control.stergm(), verbose = FALSE, ..., SAN.offsets = NULL )
nw |
A
|
formation, dissolution
|
One-sided |
constraints |
A formula specifying one or more constraints
on the support of the distribution of the networks being modeled. Multiple constraints
may be given, separated by “+” and “-” operators. See
The default is to have no constraints except those provided through
the Together with the model terms in the formula and the reference measure, the constraints define the distribution of networks being modeled. It is also possible to specify a proposal function directly either
by passing a string with the function's name (in which case,
arguments to the proposal should be specified through the
Note that not all possible combinations of constraints and reference measures are supported. However, for relatively simple constraints (i.e., those that simply permit or forbid specific dyads or sets of dyads from changing), arbitrary combinations should be possible. |
estimate |
One of "EGMME" for Equilibrium Generalized Method of Moments Estimation, based on a single network with some temporal information and making an assumption that it is a product of a STERGM process running to its stationary (equilibrium) distribution; "CMLE" for Conditional Maximum Likelihood Estimation, modeling a transition between two networks, or "CMPLE" for Conditional Maximum PseudoLikelihood Estimation, using MPLE instead of MLE. CMPLE is extremely inaccurate at this time. |
times |
For CMLE and CMPLE estimation, times or indexes at
which the networks whose transition is to be modeled are
observed. Default to |
offset.coef.form |
Numeric vector to specify offset formation parameters. |
offset.coef.diss |
Numeric vector to specify offset dissolution parameters. |
targets |
One-sided |
target.stats |
A vector specifying the values of the |
eval.loglik |
Whether or not to calculate the log-likelihood
of a CMLE STERGM fit. See |
control |
A list of control parameters for algorithm tuning.
Constructed using |
verbose |
A logical or an integer to control the amount of
progress and diagnostic information to be printed. |
... |
Additional arguments, to be passed to lower-level functions. |
SAN.offsets |
Offset coefficients (if any) to use during the SAN run. |
The stergm function uses a pair of formulas, formation and
dissolution to model tie-dynamics. The dissolution formula, however, is
parameterized in terms of tie persistence: negative coefficients imply lower
rates of persistence and postive coefficients imply higher rates.
The dissolution effects are simply the negation of these coefficients, but
the discrepancy between the terminology and interpretation has always been
unfortunate, and we have fixed this in the new tergm function.
If you are making the transition from old stergm to new tergm, note that
the dissolution formula in stergm maps to the new Persist()
operator in the tergm function, NOT the Diss() operator.
stergm() returns an object of class tergm;
see tergm() for details and methods.
Krivitsky P.N. and Handcock M.S. (2014) A Separable Model for Dynamic Networks. Journal of the Royal Statistical Society, Series B, 76(1): 29-46. doi:10.1111/rssb.12014
Krivitsky, P.N. (2012). Modeling of Dynamic Networks based on Egocentric Data with Durational Information. Pennsylvania State University Department of Statistics Technical Report, 2012(2012-01). https://web.archive.org/web/20170830053722/https://stat.psu.edu/research/technical-report-files/2012-technical-reports/TR1201A.pdf
ergm(), network, %v%, %n%, ergmTerm
A method for summary_formula() to calculate the
specified statistics for an observed networkDynamic at the
specified time point(s). See ergmTerm for more information
on the statistics that may be specified.
## S3 method for class 'networkDynamic' summary_formula(object, at, ..., basis = NULL)## S3 method for class 'networkDynamic' summary_formula(object, at, ..., basis = NULL)
object |
An |
at |
A vector of time points at which to calculate the statistics. |
... |
Further arguments passed to or used by methods. |
basis |
An optional |
A matrix with length(at) rows, one for each time
point in at, and columns for each term of the formula,
containing the corresponding statistics measured on the network.
ergm(), networkDynamic, ergmTerm,
summary.formula()
# create a toy dynamic network my.nD <- network.initialize(100,directed=FALSE) activate.vertices(my.nD, onset=0, terminus = 10) add.edges.active(my.nD,tail=1:2,head=2:3,onset=5,terminus=8) # use a summary formula to display number of isolates and edges # at discrete time points summary(my.nD~isolates+edges, at=1:10)# create a toy dynamic network my.nD <- network.initialize(100,directed=FALSE) activate.vertices(my.nD, onset=0, terminus = 10) add.edges.active(my.nD,tail=1:2,head=2:3,onset=5,terminus=8) # use a summary formula to display number of isolates and edges # at discrete time points summary(my.nD~isolates+edges, at=1:10)
tergm() fits Temporal ERGMs' (TERGMs) and Separable Temporal ERGMs' (STERGMs)
Conditional MLE (CMLE) (Krivitsky and Handcock, 2010) and Equilibrium
Generalized Method of Moments Estimator (EGMME) (Krivitsky, 2009).
tergm( formula, constraints = ~., estimate, times = NULL, offset.coef = NULL, targets = NULL, target.stats = NULL, SAN.offsets = NULL, eval.loglik = NVL(getOption("tergm.eval.loglik"), getOption("ergm.eval.loglik")), control = control.tergm(), verbose = FALSE, ..., basis = eval_lhs.formula(formula) )tergm( formula, constraints = ~., estimate, times = NULL, offset.coef = NULL, targets = NULL, target.stats = NULL, SAN.offsets = NULL, eval.loglik = NVL(getOption("tergm.eval.loglik"), getOption("ergm.eval.loglik")), control = control.tergm(), verbose = FALSE, ..., basis = eval_lhs.formula(formula) )
formula |
an ERGM formula. |
constraints |
A formula specifying one or more constraints
on the support of the distribution of the networks being modeled. Multiple constraints
may be given, separated by “+” and “-” operators. See
The default is to have no constraints except those provided through
the Together with the model terms in the formula and the reference measure, the constraints define the distribution of networks being modeled. It is also possible to specify a proposal function directly either
by passing a string with the function's name (in which case,
arguments to the proposal should be specified through the
Note that not all possible combinations of constraints and reference measures are supported. However, for relatively simple constraints (i.e., those that simply permit or forbid specific dyads or sets of dyads from changing), arbitrary combinations should be possible. |
estimate |
One of "EGMME" for Equilibrium Generalized Method of Moments Estimation, based on a single network with some temporal information and making an assumption that it is a product of a TERGM process running to its stationary (equilibrium) distribution; "CMLE" for Conditional Maximum Likelihood Estimation, modeling a transition between two networks, or "CMPLE" for Conditional Maximum PseudoLikelihood Estimation, using MPLE instead of MLE. CMPLE is extremely inaccurate at this time. |
times |
For CMLE and CMPLE estimation, times or indexes at
which the networks whose transition is to be modeled are
observed. This argument is mandatory if |
offset.coef |
Numeric vector to specify offset parameters. |
targets |
One-sided |
target.stats |
A vector specifying the values of the |
SAN.offsets |
Offset coefficients (if any) to use during the SAN run. |
eval.loglik |
Whether or not to calculate the log-likelihood
of a CMLE TERGM fit. See |
control |
A list of control parameters for algorithm tuning.
Constructed using |
verbose |
A logical or an integer to control the amount of
progress and diagnostic information to be printed. |
... |
Additional arguments, to be passed to lower-level functions. |
basis |
optional network data overriding the left hand side of |
tergm() returns an object of class tergm that
inherits from ergm and has the usual methods (coef.ergm(),
summary.ergm(), mcmc.diagnostics(), etc.) implemented for
it. Note that gof() only works for the CMLE method.
Krackhardt, D and Handcock, MS (2006) Heider vs Simmel: Emergent features in dynamic structures. ICML Workshop on Statistical Network Analysis. Springer, Berlin, Heidelberg, 2006.
Hanneke S, Fu W, and Xing EP (2010). Discrete Temporal Models of Social Networks. Electronic Journal of Statistics, 2010, 4, 585-605. doi:10.1214/09-EJS548
Krivitsky P.N. and Handcock M.S. (2014) A Separable Model for Dynamic Networks. Journal of the Royal Statistical Society, Series B, 76(1): 29-46. doi:10.1111/rssb.12014
Krivitsky, P.N. (2012). Modeling of Dynamic Networks based on Egocentric Data with Durational Information. Pennsylvania State University Department of Statistics Technical Report, 2012(2012-01). https://arxiv.org/abs/2203.06866
network, networkDynamic, and NetSeries() for the data structures,
ergm() and ergmTerm for model specification,
package vignette browseVignettes(package='tergm') for a
short demonstration, the Statnet web site
https://statnet.org/workshop-tergm/ for a tutorial
## Not run: # EGMME Example par(ask=FALSE) n<-30 g0<-network.initialize(n,dir=FALSE) # edges, degree(1), mean.age target.stats<-c( n*1/2, n*0.6, 20) dynfit<-tergm(g0 ~ Form(~edges + degree(1)) + Diss(~edges), targets = ~edges+degree(1)+mean.age, target.stats=target.stats, estimate="EGMME", control=control.tergm(SA.plot.progress=TRUE)) par(ask=TRUE) mcmc.diagnostics(dynfit) summary(dynfit) ## End(Not run) # CMLE Example data(samplk) # Fit a transition from Time 1 to Time 2 samplk12 <- tergm(list(samplk1, samplk2)~ Form(~edges+mutual+transitiveties+cyclicalties)+ Diss(~edges+mutual+transitiveties+cyclicalties), estimate="CMLE") mcmc.diagnostics(samplk12) summary(samplk12) samplk12.gof <- gof(samplk12) samplk12.gof plot(samplk12.gof) plot(samplk12.gof, plotlogodds=TRUE) # Fit a transition from Time 1 to Time 2 and from Time 2 to Time 3 jointly samplk123 <- tergm(list(samplk1, samplk2, samplk3)~ Form(~edges+mutual+transitiveties+cyclicalties)+ Diss(~edges+mutual+transitiveties+cyclicalties), estimate="CMLE") mcmc.diagnostics(samplk123) summary(samplk123)## Not run: # EGMME Example par(ask=FALSE) n<-30 g0<-network.initialize(n,dir=FALSE) # edges, degree(1), mean.age target.stats<-c( n*1/2, n*0.6, 20) dynfit<-tergm(g0 ~ Form(~edges + degree(1)) + Diss(~edges), targets = ~edges+degree(1)+mean.age, target.stats=target.stats, estimate="EGMME", control=control.tergm(SA.plot.progress=TRUE)) par(ask=TRUE) mcmc.diagnostics(dynfit) summary(dynfit) ## End(Not run) # CMLE Example data(samplk) # Fit a transition from Time 1 to Time 2 samplk12 <- tergm(list(samplk1, samplk2)~ Form(~edges+mutual+transitiveties+cyclicalties)+ Diss(~edges+mutual+transitiveties+cyclicalties), estimate="CMLE") mcmc.diagnostics(samplk12) summary(samplk12) samplk12.gof <- gof(samplk12) samplk12.gof plot(samplk12.gof) plot(samplk12.gof, plotlogodds=TRUE) # Fit a transition from Time 1 to Time 2 and from Time 2 to Time 3 jointly samplk123 <- tergm(list(samplk1, samplk2, samplk3)~ Form(~edges+mutual+transitiveties+cyclicalties)+ Diss(~edges+mutual+transitiveties+cyclicalties), estimate="CMLE") mcmc.diagnostics(samplk123) summary(samplk123)
Gives the network a series of timed proposals it can't refuse. Returns the statistics of the network, and, optionally, the final network.
tergm.godfather( formula, changes = NULL, toggles = changes[, -4, drop = FALSE], start = NULL, end = NULL, end.network = FALSE, stats.start = FALSE, verbose = FALSE, control = control.tergm.godfather() )tergm.godfather( formula, changes = NULL, toggles = changes[, -4, drop = FALSE], start = NULL, end = NULL, end.network = FALSE, stats.start = FALSE, verbose = FALSE, control = control.tergm.godfather() )
formula |
An |
changes |
A matrix with four columns: time, tail, head, and
new value, describing the changes to be made. Can only be used if
LHS of |
toggles |
A matrix with three columns: time, tail, and head,
giving the dyads which had changed. Can only be used if LHS of
|
start |
Time from which to start applying changes. Note that
the first set of changes will take effect at |
end |
Time at which to finish applying changes. Defaults to the last time point at which a change occurs. |
end.network |
Whether to return the network that
results. Defaults to |
stats.start |
Whether to return the network statistics at
|
verbose |
A logical or an integer to control the amount of
progress and diagnostic information to be printed. |
control |
A control list generated by
|
If end.network is FALSE (the default), an
mcmc object with the requested network statistics
associated with the network series produced by applying the
specified changes. Its mcmc attributes encode the
timing information: so start(out) gives the time
point associated with the first row returned, and
end(out) out the last. The "thinning interval" is
always 1.
If end.network is TRUE, return a network object with
lasttoggle "extension", representing the final network, with a
matrix of statistics described in the previous paragraph attached to it as
an attr-style attribute "stats".
simulate.tergm(), simulate_formula.network(), simulate_formula.networkDynamic()