CID-gibbs: CIDnetworks: Bayesian Inference

In bdabbs13/CIDnetworks: Generative Models for Complex Networks with Conditionally Independent Dyadic Structure

Description

Usage of CID Gibbs Samplers.

Usage

  CID.Gibbs (input,
             outcome,
             node.names,
             components,
             class.outcome="ordinal",
             fill.in.missing.edges = missing(outcome),
             new.chain = FALSE,

             draws = 100,
             burnin = -1,
             thin = 10,
             report = 100,
             auto.converge = FALSE,
             extend.max=10,
             extend.count=100,
             verbose=2,
             ...)

  ## S3 method for class 'CID.Gibbs'
 print(x, ...)
  ## S3 method for class 'CID.Gibbs'
 summary(object, ...)
  ## S3 method for class 'CID.Gibbs'
 plot(x, ...)
  ## S3 method for class 'summary.CID.Gibbs'
 print(x, ...)

  likelihood.plot(x, ...)
  intercept.plot(x, mode = c("standard","trace"), ...)
  COV.plot(x, mode = c("standard","trace","scatterplot"), ...)
  LSM.plot(x, ...)
  SBM.plot(x, ...)
  MMSBM.plot(x, ...)
  SR.plot(x, ...)

  network.plot (x, fitted.values=FALSE, ...)
  sociogram.plot (x, component.color=0, vertexcolor, add.labels = TRUE, ...)

  n.nodes(object)
  edge.list(object)
  is.net.directed(object)
  net.density(object)
  outcome(object)
  node.names(object)
  inDegree(object)
  outDegree(object)
  socio(object)
  value.mat(CID.Gibbs.object, prob = TRUE)
  value.mat.mean(object, prob = TRUE)
  switcheroo(CID.Gibbs.object)  
 
  post.pred.mat(object)
  post.pred.apply(object, FUN)
  post.pred.mean(object)
  getModeLS(object, gridsize = 50)

  

Arguments

input	An object containing information about the edges in a network. Must be one of the following classes: Matrix, CIDnetwork, or CID.Gibbs. If input is a square matrix, it is assumed to be a sociomatrix. Otherwise a matrix with 2 columns and a number of rows equal to the number of edges is required. Providing a CIDnetwork will use the associated edge.list. Providing a CID.Gibbs object will continue the MCMC chain from the last draw.
outcome	If an edgelist is provided as input and outcome is missing, the edges provided are assumed to be the ones in a binary sociomatrix. Otherwise, an outcome value must be specified for each edge in the edgelist, and any edges not provided are assumed to have no data.
node.names	Names labeling each node in the network.
CID.Gibbs.object	If desired, an existing CID.Gibbs output object can be loaded instead of a new network specification.
components	A list of sub-components, including (COV, HBM, LSM, LVM, MMSBM, SBM, SR).
class.outcome	One of "ordinal" (default, values from 0 to higher integers), "binary" (ordinal in 0 and 1) or "gaussian" (unbounded continuous values). Class is auto-detected if NULL remains in place.
fill.in.missing.edges	If TRUE, the edge list will be augmented with zeroes for all unspecified but possible edges. By default, if an outcome is specified, these edges will not be added.
new.chain	If a CID.Gibbs object is provided, the default value of FALSE will return both the old and new MCMC chain combined. A value of TRUE will drop the old chain completely.
draws	Number of draws to return.
burnin	Number of draws to burnin. A negative value will automatically determine burnin amount.
thin	Amount of draws to thin the chain by.
report	Number of draws between reporting total draws so far.
auto.converge	When true, a Geweke convergence test on log-likleihood to detect convergence.
extend.max	Maximum number of times chain will be extended until it returns without converging.
extend.count	Number of draws to extend chain by if convergence test fails
verbose	Level of output to be displayed while running. A value of 0 will return little or no output. A value of 1 will only notify of warnings of misuse. A value of 2 will report progress of MCMC chain. A value of 3 or higher will report debugging values.
...	Further arguments to be passed to the Gibbs sampler routine or the plot routine. See details for more.
x, object	An object outputted from CID.Gibbs.
mode	Controls which diagnostic plot is made.
fitted.values	If TRUE, plots the fitted tie strength under the Gibbs sampler. If FALSE, plots the network outcomes as entered.
component.color	If non-zero, colors the nodes in the sociogram according to the output of the Gibbs sampler.
vertexcolor	User-passed vertex colors for sociogram.plot .
add.labels	When true, node labels are included on nodes.
trace	If selected, displays the Gibbs sampler trace plot for the intercept rather than a point and interval.
prob	In value.mat, converts the linear predictor value to the probability of a binary edge.
FUN	function to apply to all posterior samples.
gridsize	Gridsize for evaluating posterior distribution and searching for mode.

Details

This is the main routine for running a Gibbs sampler on any of the CID models. See the vignettes for more information.

Value

CID.Gibbs outputs a list containing a CID object, the results of the Gibbs sampler, and the Deviance Information Criterion estimate for the Gibbs.

Author(s)

A.C. Thomas <[e-mail]>

bdabbs13/CIDnetworks documentation built on Nov. 15, 2019, 2:41 a.m.