Usage
gergm(formula, covariate_data = NULL, normalization_type = c("log",
"division"), network_is_directed = c(TRUE, FALSE),
use_MPLE_only = c(FALSE, TRUE), transformation_type = c("Cauchy",
"LogCauchy", "Gaussian", "LogNormal"), estimation_method = c("Gibbs",
"Metropolis"), maximum_number_of_lambda_updates = 10,
maximum_number_of_theta_updates = 10,
number_of_networks_to_simulate = 500, thin = 1, proposal_variance = 0.1,
downweight_statistics_together = TRUE, MCMC_burnin = 100, seed = 123,
convergence_tolerance = 0.01, MPLE_gain_factor = 0,
acceptable_fit_p_value_threshold = 0.05, force_x_theta_updates = 1,
force_x_lambda_updates = 1, output_directory = NULL, output_name = NULL,
generate_plots = TRUE, verbose = TRUE, omit_intercept_term = FALSE, ...)Arguments
formula
A formula object that specifies the relationship between
statistics and the observed network. Currently, the user may specify a model
using any combination of the following statistics: `out2stars(alpha = 1)`,
`in2stars(alpha = 1)`, `ctriads(alpha = 1)`,
covariate_data
A data frame containing node level covariates the user
wished to transform into sender or reciever effects. It must have row names
that match every entry in colnames(raw_network), should have descriptive
column names. If left NULL, then no sender or reci
normalization_type
If only a raw_network is provided and
omit_intercept_term = TRUE then, the function
will automatically check to determine if all edges fall in the [0,1] interval.
If edges are determined to fall outside of this interval, then a trasformation
onto the inte
network_is_directed
Logical specifying whether or not the observed
network is directed. Default is TRUE.
use_MPLE_only
Logical specifying whether or not only the maximum pseudo
likelihood estimates should be obtained. In this case, no simulations will be
performed. Default is FALSE.
transformation_type
Specifies how covariates are transformed onto the
raw network. When working with heavly tailed data that are not strictly
positive, select "Cauchy" to transform the data using a Cauchy distribution.
If data are strictly positive and heavy tailed (such as
estimation_method
Simulation method for MCMC estimation. Default is
"Gibbs" which will generally be faster with well behaved networks but will not
allow for exponential downweighting.
maximum_number_of_lambda_updates
Maximum number of iterations of outer
MCMC loop which alternately estimates transform parameters and ERGM
parameters. In the case that data_transformation = NULL, this argument is
ignored. Default is 10.
maximum_number_of_theta_updates
Maximum number of iterations within the
MCMC inner loop which estimates the ERGM parameters. Default is 100.
number_of_networks_to_simulate
Number of simulations generated for
estimation via MCMC. Default is 500.
thin
The proportion of samples that are kept from each simulation. For
example, thin = 1/200 will keep every 200th network in the overall simulated
sample. Default is 1.
proposal_variance
The variance specified for the Metropolis Hastings
simulation method. This parameter is inversely proportional to the average
acceptance rate of the M-H sampler and should be adjusted so that the average
acceptance rate is approximately 0.25. Default is 0
downweight_statistics_together
Logical specifying whether or not the
weights should be applied inside or outside the sum. Default is TRUE and user
should not select FALSE under normal circumstances.
MCMC_burnin
Number of samples from the MCMC simulation procedure that
will be discarded before drawing the samples used for estimation.
Default is 100.
seed
Seed used for reproducibility. Default is 123.
convergence_tolerance
Threshold designated for stopping criterion. If
the difference of parameter estimates from one iteration to the next all have
a p -value (under a paired t-test) greater than this value, the parameter
estimates are declared to have converged. Default is 0.
MPLE_gain_factor
Multiplicative constant between 0 and 1 that controls
how far away the initial theta estimates will be from the standard MPLEs via
a one step Fisher update. In the case of strongly dependent data, it is
suggested to use a value of 0.10. Default is 0.
acceptable_fit_p_value_threshold
A p-value threshold for how closely
statistics of observed network conform to statistics of networks simulated
from GERGM parameterized by converged final parameter estimates. Default value
is 0.05.
force_x_theta_updates
Defaults to 1 where theta estimation is not
allowed to converge until thetas have updated for x iterations . Useful when
model is not degenerate but simulated statistics do not match observed network
well when algorithm stops after first y updates.
force_x_lambda_updates
Defaults to 1 where lambda estimation is not
allowed to converge until lambdas have updated for x iterations . Useful when
model is not degenerate but simulated statistics do not match observed network
well when algorithm stops after first y updates.
output_directory
The directory where you would like output generated
by the GERGM estimation proceedure to be saved (if output_name is specified).
This includes, GOF, trace, and parameter estimate plots, as well as a summary
of the estimation proceedure and an .Rdata file
output_name
The common name stem you would like to assign to all
objects output by the gergm function. Default value of NULL will not save any
output directly to .pdf files, it will be printed to the console instead. Must
be a character string or NULL. For example, i
generate_plots
Defaults to TRUE, if FALSE, then no diagnostic or
parameter plots are generated.
verbose
Defaults to TRUE (providing lots of output while model is
running). Can be set to FALSE if the user wishes to see less output.
omit_intercept_term
Defualts to FALSE, can be set to TRUE if the
user wishes to omit the model intercept term.
...
Optional arguments, currently unsupported.