Type:	Package
Title:	Handling Missing Data in Stochastic Block Models
Version:	1.0.5
Maintainer:	Julien Chiquet <julien.chiquet@inrae.fr>
Description:	When a network is partially observed (here, NAs in the adjacency matrix rather than 1 or 0 due to missing information between node pairs), it is possible to account for the underlying process that generates those NAs. 'missSBM', presented in 'Barbillon, Chiquet and Tabouy' (2022) <doi:10.18637/jss.v101.i12>, adjusts the popular stochastic block model from network data sampled under various missing data conditions, as described in 'Tabouy, Barbillon and Chiquet' (2019) <doi:10.1080/01621459.2018.1562934>.
URL:	https://grosssbm.github.io/missSBM/
BugReports:	https://github.com/grossSBM/missSBM/issues
License:	GPL-3
Encoding:	UTF-8
LazyData:	true
RoxygenNote:	7.3.2
Depends:	R (≥ 3.4.0)
Imports:	Rcpp, methods, igraph, nloptr, ggplot2, future.apply, R6, rlang, sbm, magrittr, Matrix, RSpectra
LinkingTo:	Rcpp, RcppArmadillo, nloptr
Collate:	'utils_missSBM.R' 'R6Class-networkSampling.R' 'R6Class-networkSampling_fit.R' 'R6Class-simpleSBM_fit.R' 'R6Class-missSBM_fit.R' 'R6Class-missSBM_collection.R' 'R6Class-networkSampler.R' 'R6Class-partlyObservedNetwork.R' 'RcppExports.R' 'er_network.R' 'estimateMissSBM.R' 'frenchblog2007.R' 'kmeans.R' 'missSBM-package.R' 'observeNetwork.R' 'war.R'
Suggests:	aricode, blockmodels, corrplot, future, testthat (≥ 2.1.0), covr, knitr, rmarkdown, spelling
VignetteBuilder:	knitr
Language:	en-US
NeedsCompilation:	yes
Packaged:	2025-03-13 08:07:21 UTC; jchiquet
Author:	Julien Chiquet [aut, cre], Pierre Barbillon [aut], Timothée Tabouy [aut], Jean-Benoist Léger [ctb] (provided C++ implementaion of K-means), François Gindraud [ctb] (provided C++ interface to NLopt), großBM team [ctb]
Repository:	CRAN
Date/Publication:	2025-03-13 08:30:02 UTC

missSBM: Handling Missing Data in Stochastic Block Models

Description

When a network is partially observed (here, NAs in the adjacency matrix rather than 1 or 0 due to missing information between node pairs), it is possible to account for the underlying process that generates those NAs. 'missSBM', presented in 'Barbillon, Chiquet and Tabouy' (2022) doi:10.18637/jss.v101.i12, adjusts the popular stochastic block model from network data sampled under various missing data conditions, as described in 'Tabouy, Barbillon and Chiquet' (2019) doi:10.1080/01621459.2018.1562934.

The missSBM package provides the following top-level functions functions:

Details

• observeNetwork a function to draw a partially observe network from an existing, fully observed network according to a variety of sampling designs

• estimateMissSBM a function to perform inference of SBM from a partially observed under various sampling designs.

These function leads to the manipulation of a variety of R objects instantiated from some R6 classes, with their respective fields and methods. They are all generated by the top-level functions itemized above, so that the user should generally not use their constructor or internal methods directly. The user should only have a basic understanding of the fields of each object to manipulate the output in R. The main objects are the following:

• missSBM_fit an object that put together an SBM fit and and network sampling fit - the main point of the missSBM package !

• missSBM_collection an object to store a collection of missSBM_fit, ordered by number of block

• SimpleSBM_fit_MNAR an object to define and store an SBM fit with MNAR values

• SimpleSBM_fit_noCov an object to define and store an SBM fit without covariate, MAR values

• SimpleSBM_fit_withCov an object to define and store an SBM fit with covariates, MAR values

• networkSampling an object to define and store a network sampling fit

missSBM extends some functionality of the package sbm, by inheriting from classes and methods associated to simple stochastic block models.

Author(s)

Maintainer: Julien Chiquet julien.chiquet@inrae.fr (ORCID)

Authors:

• Pierre Barbillon pierre.barbillon@agroparistech.fr (ORCID)

• Timothée Tabouy timothee.tabouy@agroparistech.fr

Other contributors:

• Jean-Benoist Léger jbleger@hds.utc.fr (provided C++ implementaion of K-means) [contributor]

• François Gindraud francois.gindraud@gmail.com (provided C++ interface to NLopt) [contributor]

• großBM team [contributor]

Pierre Barbillon pierre.barbillon@agroparistech.fr

Julien Chiquet julien.chiquet@inrae.fr

Timothée Tabouy timothee.tabouy@gmail.com

References

• Pierre Barbillon, Julien Chiquet & Timothée Tabouy (2022) "missSBM: An R Package for Handling Missing Values in the Stochastic Block Model", Journal of Statistical Software, doi:10.18637/jss.v101.i12

• Timothée Tabouy, Pierre Barbillon & Julien Chiquet (2019) “Variational Inference for Stochastic Block Models from Sampled Data”, Journal of the American Statistical Association, doi:10.1080/01621459.2018.1562934

See Also

Useful links:

• https://grosssbm.github.io/missSBM/

• Report bugs at https://github.com/grossSBM/missSBM/issues

Pipe operator

Description

See magrittr::%>% for details.

Usage

lhs %>% rhs

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

Description

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

Details

It is not designed not be call by the user

Super classes

sbm::SBM -> sbm::SimpleSBM -> SimpleSBM_fit

Active bindings

Methods

Public methods

• SimpleSBM_fit$new()

• SimpleSBM_fit$doVEM()

• SimpleSBM_fit$reorder()

• SimpleSBM_fit$clone()

Method new()

constructor for simpleSBM_fit for missSBM purpose

Usage

SimpleSBM_fit$new(networkData, clusterInit, covarList = list())

Arguments

Method doVEM()

method to perform estimation via variational EM

Usage

SimpleSBM_fit$doVEM(
  threshold = 0.01,
  maxIter = 100,
  fixPointIter = 3,
  trace = FALSE
)

Arguments

Method reorder()

permute group labels by order of decreasing probability

Usage

SimpleSBM_fit$reorder()

Method clone()

The objects of this class are cloneable with this method.

Usage

SimpleSBM_fit$clone(deep = FALSE)

Arguments

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

Description

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

Details

It is not designed not be call by the user

Super classes

sbm::SBM -> sbm::SimpleSBM -> missSBM::SimpleSBM_fit -> missSBM::SimpleSBM_fit_noCov -> SimpleSBM_MNAR_noCov

Active bindings

Methods

Public methods

• SimpleSBM_fit_MNAR$new()

• SimpleSBM_fit_MNAR$update_parameters()

• SimpleSBM_fit_MNAR$update_blocks()

• SimpleSBM_fit_MNAR$clone()

Method new()

constructor for simpleSBM_fit for missSBM purpose

Usage

SimpleSBM_fit_MNAR$new(networkData, clusterInit)

Arguments

Method update_parameters()

update parameters estimation (M-step)

Usage

SimpleSBM_fit_MNAR$update_parameters(nu = NULL)

Arguments

Method update_blocks()

update variational estimation of blocks (VE-step)

Usage

SimpleSBM_fit_MNAR$update_blocks(log_lambda = 0)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

SimpleSBM_fit_MNAR$clone(deep = FALSE)

Arguments

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

Description

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

Details

It is not designed not be call by the user

Super classes

sbm::SBM -> sbm::SimpleSBM -> missSBM::SimpleSBM_fit -> SimpleSBM_fit_noCov

Active bindings

Methods

Public methods

• SimpleSBM_fit_noCov$update_parameters()

• SimpleSBM_fit_noCov$update_blocks()

• SimpleSBM_fit_noCov$clone()

Method update_parameters()

update parameters estimation (M-step)

Usage

SimpleSBM_fit_noCov$update_parameters(...)

Arguments

Method update_blocks()

update variational estimation of blocks (VE-step)

Usage

SimpleSBM_fit_noCov$update_blocks(...)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

SimpleSBM_fit_noCov$clone(deep = FALSE)

Arguments

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

Description

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

This internal class is designed to adjust a binary Stochastic Block Model in the context of missSBM.

Details

It is not designed not be call by the user

Super classes

sbm::SBM -> sbm::SimpleSBM -> missSBM::SimpleSBM_fit -> SimpleSBM_fit_withCov

Active bindings

Methods

Public methods

• SimpleSBM_fit_withCov$update_parameters()

• SimpleSBM_fit_withCov$update_blocks()

• SimpleSBM_fit_withCov$clone()

Method update_parameters()

update parameters estimation (M-step)

Usage

SimpleSBM_fit_withCov$update_parameters(...)

Arguments

Method update_blocks()

update variational estimation of blocks (VE-step)

Usage

SimpleSBM_fit_withCov$update_blocks(...)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

SimpleSBM_fit_withCov$clone(deep = FALSE)

Arguments

Class for defining a block dyad sampler

Description

Class for defining a block dyad sampler

Class for defining a block dyad sampler

Super classes

missSBM::networkSampling -> missSBM::networkSampler -> missSBM::dyadSampler -> blockDyadSampler

Active bindings

Methods

Public methods

• blockDyadSampler$new()

• blockDyadSampler$clone()

Method new()

constructor for networkSampling

Usage

blockDyadSampler$new(
  parameters = NA,
  nbNodes = NA,
  directed = FALSE,
  clusters = NA
)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

blockDyadSampler$clone(deep = FALSE)

Arguments

Class for fitting a block-dyad sampling

Description

Class for fitting a block-dyad sampling

Class for fitting a block-dyad sampling

Super classes

missSBM::networkSampling -> missSBM::networkSamplingDyads_fit -> blockDyadSampling_fit

Active bindings

Methods

Public methods

• blockDyadSampling_fit$new()

• blockDyadSampling_fit$update_parameters()

• blockDyadSampling_fit$clone()

Method new()

constructor

Usage

blockDyadSampling_fit$new(partlyObservedNetwork, blockInit)

Arguments

Method update_parameters()

a method to update the estimation of the parameters. By default, nothing to do (corresponds to MAR sampling)

Usage

blockDyadSampling_fit$update_parameters(nu, Z)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

blockDyadSampling_fit$clone(deep = FALSE)

Arguments

Class for defining a block node sampler

Description

Class for defining a block node sampler

Class for defining a block node sampler

Super classes

missSBM::networkSampling -> missSBM::networkSampler -> missSBM::nodeSampler -> blockNodeSampler

Methods

Public methods

• blockNodeSampler$new()

• blockNodeSampler$clone()

Method new()

constructor for networkSampling

Usage

blockNodeSampler$new(
  parameters = NA,
  nbNodes = NA,
  directed = FALSE,
  clusters = NA
)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

blockNodeSampler$clone(deep = FALSE)

Arguments

Class for fitting a block-node sampling

Description

Class for fitting a block-node sampling

Class for fitting a block-node sampling

Super classes

missSBM::networkSampling -> missSBM::networkSamplingNodes_fit -> blockNodeSampling_fit

Active bindings

Methods

Public methods

• blockNodeSampling_fit$new()

• blockNodeSampling_fit$update_parameters()

• blockNodeSampling_fit$clone()

Method new()

constructor

Usage

blockNodeSampling_fit$new(partlyObservedNetwork, blockInit)

Arguments

Method update_parameters()

a method to update the estimation of the parameters. By default, nothing to do (corresponds to MAR sampling)

Usage

blockNodeSampling_fit$update_parameters(imputedNet, Z)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

blockNodeSampling_fit$clone(deep = FALSE)

Arguments

Extract model coefficients

Description

Extracts model coefficients from objects missSBM_fit returned by estimateMissSBM()

Usage

## S3 method for class 'missSBM_fit'
coef(
  object,
  type = c("mixture", "connectivity", "covariates", "sampling"),
  ...
)

Arguments

Value

A vector or matrix of coefficients extracted from the missSBM_fit model.

Class for fitting a dyad sampling with covariates

Description

Class for fitting a dyad sampling with covariates

Class for fitting a dyad sampling with covariates

Super classes

missSBM::networkSampling -> missSBM::networkSamplingDyads_fit -> covarDyadSampling_fit

Active bindings

Methods

Public methods

• covarDyadSampling_fit$new()

• covarDyadSampling_fit$clone()

Method new()

constructor

Usage

covarDyadSampling_fit$new(partialNet, ...)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

covarDyadSampling_fit$clone(deep = FALSE)

Arguments

Class for fitting a node-centered sampling with covariate

Description

Class for fitting a node-centered sampling with covariate

Class for fitting a node-centered sampling with covariate

Super classes

missSBM::networkSampling -> missSBM::networkSamplingNodes_fit -> covarNodeSampling_fit

Active bindings

Methods

Public methods

• covarNodeSampling_fit$new()

• covarNodeSampling_fit$clone()

Method new()

constructor

Usage

covarNodeSampling_fit$new(partlyObservedNetwork, ...)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

covarNodeSampling_fit$clone(deep = FALSE)

Arguments

Class for defining a degree sampler

Description

Class for defining a degree sampler

Class for defining a degree sampler

Super classes

missSBM::networkSampling -> missSBM::networkSampler -> missSBM::nodeSampler -> degreeSampler

Methods

Public methods

• degreeSampler$new()

• degreeSampler$clone()

Method new()

constructor for networkSampling

Usage

degreeSampler$new(parameters = NA, degrees = NA, directed = FALSE)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

degreeSampler$clone(deep = FALSE)

Arguments

Class for fitting a degree sampling

Description

Class for fitting a degree sampling

Class for fitting a degree sampling

Super classes

missSBM::networkSampling -> missSBM::networkSamplingNodes_fit -> degreeSampling_fit

Active bindings

Methods

Public methods

• degreeSampling_fit$new()

• degreeSampling_fit$update_parameters()

• degreeSampling_fit$update_imputation()

• degreeSampling_fit$clone()

Method new()

constructor

Usage

degreeSampling_fit$new(partlyObservedNetwork, blockInit, connectInit)

Arguments

Method update_parameters()

a method to update the estimation of the parameters. By default, nothing to do (corresponds to MAR sampling)

Usage

degreeSampling_fit$update_parameters(imputedNet, ...)

Arguments

Method update_imputation()

a method to update the imputation of the missing entries.

Usage

degreeSampling_fit$update_imputation(PI, ...)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

degreeSampling_fit$clone(deep = FALSE)

Arguments

Class for defining a double-standard sampler

Description

Class for defining a double-standard sampler

Class for defining a double-standard sampler

Super classes

missSBM::networkSampling -> missSBM::networkSampler -> missSBM::dyadSampler -> doubleStandardSampler

Methods

Public methods

• doubleStandardSampler$new()

• doubleStandardSampler$clone()

Method new()

constructor for networkSampling

Usage

doubleStandardSampler$new(parameters = NA, adjMatrix = NA, directed = FALSE)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

doubleStandardSampler$clone(deep = FALSE)

Arguments

Class for fitting a double-standard sampling

Description

Class for fitting a double-standard sampling

Class for fitting a double-standard sampling

Super classes

missSBM::networkSampling -> missSBM::networkSamplingDyads_fit -> doubleStandardSampling_fit

Active bindings

Methods

Public methods

• doubleStandardSampling_fit$new()

• doubleStandardSampling_fit$update_parameters()

• doubleStandardSampling_fit$update_imputation()

• doubleStandardSampling_fit$clone()

Method new()

constructor

Usage

doubleStandardSampling_fit$new(partlyObservedNetwork, ...)

Arguments

Method update_parameters()

a method to update the estimation of the parameters. By default, nothing to do (corresponds to MAR sampling)

Usage

doubleStandardSampling_fit$update_parameters(nu, ...)

Arguments

Method update_imputation()

a method to update the imputation of the missing entries.

Usage

doubleStandardSampling_fit$update_imputation(nu)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

doubleStandardSampling_fit$clone(deep = FALSE)

Arguments

Virtual class for all dyad-centered samplers

Description

Virtual class for all dyad-centered samplers

Virtual class for all dyad-centered samplers

Super classes

missSBM::networkSampling -> missSBM::networkSampler -> dyadSampler

Methods

Public methods

• dyadSampler$new()

• dyadSampler$clone()

Method new()

constructor for networkSampling

Usage

dyadSampler$new(type = NA, parameters = NA, nbNodes = NA, directed = FALSE)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

dyadSampler$clone(deep = FALSE)

Arguments

Class for fitting a dyad sampling

Description

Class for fitting a dyad sampling

Class for fitting a dyad sampling

Super classes

missSBM::networkSampling -> missSBM::networkSamplingDyads_fit -> dyadSampling_fit

Active bindings

Methods

Public methods

• dyadSampling_fit$new()

• dyadSampling_fit$clone()

Method new()

constructor

Usage

dyadSampling_fit$new(partlyObservedNetwork, ...)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

dyadSampling_fit$clone(deep = FALSE)

Arguments

ER ego centered network

Description

A dataset containing the weighted PPI network centered around the ESR1 (ER) protein

Usage

er_network

Format

A sparse symmetric matrix with 741 rows and 741 columns ESR1

Source

https://string-db.org/

Examples

data("er_network")
class(er_network)

Estimation of simple SBMs with missing data

Description

Variational EM inference of Stochastic Block Models indexed by block number from a partially observed network.

Usage

estimateMissSBM(
  adjacencyMatrix,
  vBlocks,
  sampling,
  covariates = list(),
  control = list()
)

Arguments

Details

Internal functions use future_lapply, so set your plan to 'multisession' or 'multicore' to use several cores/workers. The list of parameters control tunes more advanced features, such as the initialization, how covariates are handled in the model, and the variational EM algorithm:

• useCov logical. If covariates is not null, should they be used for the for the SBM inference (or just for the sampling)? Default is TRUE.

• clusterInit Initial method for clustering: either a character ("spectral") or a list with length(vBlocks) vectors, each with size ncol(adjacencyMatrix), providing a user-defined clustering. Default is "spectral". similarity An R x R -> R function to compute similarities between node covariates. Default is l1_similarity, that is, -abs(x-y). Only relevant when the covariates are node-centered (i.e. covariates is a list of size-N vectors).

• threshold V-EM algorithm stops stop when an optimization step changes the objective function or the parameters by less than threshold. Default is 1e-2.

• maxIter V-EM algorithm stops when the number of iteration exceeds maxIter. Default is 50.

• fixPointIter number of fix-point iterations in the V-E step. Default is 3.

• exploration character indicating the kind of exploration used among "forward", "backward", "both" or "none". Default is "both".

• iterates integer for the number of iterations during exploration. Only relevant when exploration is different from "none". Default is 1.

• trace logical for verbosity. Default is TRUE.

The different sampling designs are split into two families in which we find dyad-centered and node-centered samplings. See doi:10.1080/01621459.2018.1562934 for a complete description.

• Missing at Random (MAR)

  • dyad parameter = p = Prob(Dyad(i,j) is observed)

  • node parameter = p = Prob(Node i is observed)

  • covar-dyad": parameter = beta in R^M, such that Prob(Dyad (i,j) is observed) = logistic(parameter' covarArray (i,j, .))

  • covar-node": parameter = nu in R^M such that Prob(Node i is observed) = logistic(parameter' covarMatrix (i,)

  • snowball": parameter = number of waves with Prob(Node i is observed in the 1st wave)

• Missing Not At Random (MNAR)

  • double-standard parameter = (p0,p1) with p0 = Prob(Dyad (i,j) is observed | the dyad is equal to 0), p1 = Prob(Dyad (i,j) is observed | the dyad is equal to 1)

  • block-node parameter = c(p(1),...,p(Q)) and p(q) = Prob(Node i is observed | node i is in cluster q)

  • block-dyad parameter = c(p(1,1),...,p(Q,Q)) and p(q,l) = Prob(Edge (i,j) is observed | node i is in cluster q and node j is in cluster l)

Value

Returns an R6 object with class missSBM_collection.

See Also

observeNetwork, missSBM_collection and missSBM_fit.

Examples

## SBM parameters
N <- 100 # number of nodes
Q <- 3   # number of clusters
pi <- rep(1,Q)/Q     # block proportion
theta <- list(mean = diag(.45,Q) + .05 ) # connectivity matrix

## Sampling parameters
samplingParameters <- .75 # the sampling rate
sampling  <- "dyad"      # the sampling design

## generate a undirected binary SBM with no covariate
sbm <- sbm::sampleSimpleSBM(N, pi, theta)

## Uncomment to set parallel computing with future
## future::plan("multicore", workers = 2)

## Sample some dyads data + Infer SBM with missing data
collection <-
   observeNetwork(sbm$networkData, sampling, samplingParameters) %>%
   estimateMissSBM(vBlocks = 1:4, sampling = sampling)
plot(collection, "monitoring")
plot(collection, "icl")

collection$ICL
coef(collection$bestModel$fittedSBM, "connectivity")

myModel <- collection$bestModel
plot(myModel, "expected")
plot(myModel, "imputed")
plot(myModel, "meso")
coef(myModel, "sampling")
coef(myModel, "connectivity")
predict(myModel)[1:5, 1:5]

Extract model fitted values from object missSBM_fit, return by estimateMissSBM()

Description

Extract model fitted values from object missSBM_fit, return by estimateMissSBM()

Usage

## S3 method for class 'missSBM_fit'
fitted(object, ...)

Arguments

Value

A matrix of estimated probabilities of connection

Political Blogosphere network prior to 2007 French presidential election

Description

French Political Blogosphere network dataset consists of a single day snapshot of over 200 political blogs automatically extracted the 14 October 2006 and manually classified by the "Observatoire Présidentielle" project. Originally part of the 'mixer' package

Usage

frenchblog2007

Format

An igraph object with 196 nodes. The vertex attribute "party" provides a possible clustering of the nodes.

Source

https://www.meltwater.com/en/suite/consumer-intelligence?utm_source=direct&utm_medium=linkfluence

Examples

data(frenchblog2007)
igraph::V(frenchblog2007)$party
igraph::plot.igraph(frenchblog2007,
  vertex.color = factor(igraph::V(frenchblog2007)$party),
  vertex.label = NA
 )

L1-similarity

Description

Compute l1-similarity between two vectors

Usage

l1_similarity(x, y)

Arguments

Value

a vector equal to -abs(x-y)

An R6 class to represent a collection of SBM fits with missing data

Description

The function estimateMissSBM() fits a collection of SBM with missing data for a varying number of block. These models with class missSBM_fit are stored in an instance of an object with class missSBM_collection, described here.

Fields are accessed via active binding and cannot be changed by the user.

This class comes with a set of R6 methods, some of them being useful for the user and exported as S3 methods. See the documentation for show() and print()

Active bindings

Methods

Public methods

• missSBM_collection$new()

• missSBM_collection$estimate()

• missSBM_collection$explore()

• missSBM_collection$plot()

• missSBM_collection$show()

• missSBM_collection$print()

• missSBM_collection$clone()

Method new()

constructor for networkSampling

Usage

missSBM_collection$new(partlyObservedNet, sampling, clusterInit, control)

Arguments

Method estimate()

method to launch the estimation of the collection of models

Usage

missSBM_collection$estimate(control)

Arguments

Method explore()

method for performing exploration of the ICL

Usage

missSBM_collection$explore(control)

Arguments

Method plot()

plot method for missSBM_collection

Usage

missSBM_collection$plot(type = c("icl", "elbo", "monitoring"))

Arguments

Method show()

show method for missSBM_collection

Usage

missSBM_collection$show()

Method print()

User friendly print method

Usage

missSBM_collection$print()

Method clone()

The objects of this class are cloneable with this method.

Usage

missSBM_collection$clone(deep = FALSE)

Arguments

Examples

## Uncomment to set parallel computing with future
## future::plan("multicore", workers = 2)

## Sample 75% of dyads in  French political Blogosphere's network data
adjacencyMatrix <- missSBM::frenchblog2007 %>%
  igraph::delete.vertices(1:100) %>%
  igraph::as_adj () %>%
  missSBM::observeNetwork(sampling = "dyad", parameters = 0.75)
collection <- estimateMissSBM(adjacencyMatrix, 1:5, sampling = "dyad")
class(collection)

An R6 class to represent an SBM fit with missing data

Description

The function estimateMissSBM() fits a collection of SBM for varying number of block. Each fitted SBM is an instance of an R6 object with class missSBM_fit, described here.

Fields are accessed via active binding and cannot be changed by the user.

This class comes with a set of R6 methods, some of them being useful for the user and exported as S3 methods. See the documentation for show(), print(), fitted(), predict(), plot().

Active bindings

Methods

Public methods

• missSBM_fit$new()

• missSBM_fit$doVEM()

• missSBM_fit$show()

• missSBM_fit$print()

• missSBM_fit$clone()

Method new()

constructor for networkSampling

Usage

missSBM_fit$new(partlyObservedNet, netSampling, clusterInit, useCov = TRUE)

Arguments

Method doVEM()

a method to perform inference of the current missSBM fit with variational EM

Usage

missSBM_fit$doVEM(
  control = list(threshold = 0.01, maxIter = 100, fixPointIter = 3, trace = TRUE)
)

Arguments

Method show()

show method for missSBM_fit

Usage

missSBM_fit$show()

Method print()

User friendly print method

Usage

missSBM_fit$print()

Method clone()

The objects of this class are cloneable with this method.

Usage

missSBM_fit$clone(deep = FALSE)

Arguments

Examples

## Sample 75% of dyads in  French political Blogosphere's network data
adjMatrix <- missSBM::frenchblog2007 %>%
  igraph::as_adj (sparse = FALSE) %>%
  missSBM::observeNetwork(sampling = "dyad", parameters = 0.75)
collection <- estimateMissSBM(adjMatrix, 3:5, sampling = "dyad")
my_missSBM_fit <- collection$bestModel
class(my_missSBM_fit)
plot(my_missSBM_fit, "imputed")

Definition of R6 Class 'networkSampling_sampler'

Description

Definition of R6 Class 'networkSampling_sampler'

Definition of R6 Class 'networkSampling_sampler'

Details

This class is use to define a sampling model for a network. Inherits from 'networkSampling'. Owns a rSampling method which takes an adjacency matrix as an input and send back an object with class partlyObservedNetwork.

Super class

missSBM::networkSampling -> networkSampler

Active bindings

Methods

Public methods

• networkSampler$new()

• networkSampler$rSamplingMatrix()

• networkSampler$clone()

Method new()

constructor for networkSampling

Usage

networkSampler$new(type = NA, parameters = NA, nbNodes = NA, directed = FALSE)

Arguments

Method rSamplingMatrix()

a method for drawing a sampling matrix according to the current sampling design

Usage

networkSampler$rSamplingMatrix()

Method clone()

The objects of this class are cloneable with this method.

Usage

networkSampler$clone(deep = FALSE)

Arguments

See Also

partlyObservedNetwork

Definition of R6 Class 'networkSampling'

Description

Definition of R6 Class 'networkSampling'

Definition of R6 Class 'networkSampling'

Details

this virtual class is the mother of all subtypes of networkSampling (either sampler or fit) It is used to define a sampling model for a network. It has a rSampling method which takes an adjacency matrix as an input and send back an object with class partlyObservedNetwork.

Active bindings

Methods

Public methods

• networkSampling$new()

• networkSampling$show()

• networkSampling$print()

• networkSampling$clone()

Method new()

constructor for networkSampling

Usage

networkSampling$new(type = NA, parameters = NA)

Arguments

Method show()

show method

Usage

networkSampling$show(
  type = paste0(private$name, "-model for network sampling\n")
)

Arguments

Method print()

User friendly print method

Usage

networkSampling$print()

Method clone()

The objects of this class are cloneable with this method.

Usage

networkSampling$clone(deep = FALSE)

Arguments

Virtual class used to define a family of networkSamplingDyads_fit

Description

Virtual class used to define a family of networkSamplingDyads_fit

Virtual class used to define a family of networkSamplingDyads_fit

Super class

missSBM::networkSampling -> networkSamplingDyads_fit

Active bindings

Methods

Public methods

• networkSamplingDyads_fit$new()

• networkSamplingDyads_fit$show()

• networkSamplingDyads_fit$update_parameters()

• networkSamplingDyads_fit$update_imputation()

• networkSamplingDyads_fit$clone()

Method new()

constructor for networkSampling_fit

Usage

networkSamplingDyads_fit$new(partlyObservedNetwork, name)

Arguments

Method show()

show method

Usage

networkSamplingDyads_fit$show()

Method update_parameters()

a method to update the estimation of the parameters. By default, nothing to do (corresponds to MAR sampling)

Usage

networkSamplingDyads_fit$update_parameters(...)

Arguments

Method update_imputation()

a method to update the imputation of the missing entries.

Usage

networkSamplingDyads_fit$update_imputation(nu)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

networkSamplingDyads_fit$clone(deep = FALSE)

Arguments

Virtual class used to define a family of networkSamplingNodes_fit

Description

Virtual class used to define a family of networkSamplingNodes_fit

Virtual class used to define a family of networkSamplingNodes_fit

Super class

missSBM::networkSampling -> networkSamplingNodes_fit

Active bindings

Methods

Public methods

• networkSamplingNodes_fit$new()

• networkSamplingNodes_fit$show()

• networkSamplingNodes_fit$update_parameters()

• networkSamplingNodes_fit$update_imputation()

• networkSamplingNodes_fit$clone()

Method new()

constructor

Usage

networkSamplingNodes_fit$new(partlyObservedNetwork, name)

Arguments

Method show()

show method

Usage

networkSamplingNodes_fit$show()

Method update_parameters()

a method to update the estimation of the parameters. By default, nothing to do (corresponds to MAR sampling)

Usage

networkSamplingNodes_fit$update_parameters(...)

Arguments

Method update_imputation()

a method to update the imputation of the missing entries.

Usage

networkSamplingNodes_fit$update_imputation(nu)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

networkSamplingNodes_fit$clone(deep = FALSE)

Arguments

Virtual class for all node-centered samplers

Description

Virtual class for all node-centered samplers

Virtual class for all node-centered samplers

Super classes

missSBM::networkSampling -> missSBM::networkSampler -> nodeSampler

Methods

Public methods

• nodeSampler$clone()

Method clone()

The objects of this class are cloneable with this method.

Usage

nodeSampler$clone(deep = FALSE)

Arguments

Class for fitting a node sampling

Description

Class for fitting a node sampling

Class for fitting a node sampling

Super classes

missSBM::networkSampling -> missSBM::networkSamplingNodes_fit -> nodeSampling_fit

Active bindings

Methods

Public methods

• nodeSampling_fit$new()

• nodeSampling_fit$clone()

Method new()

constructor

Usage

nodeSampling_fit$new(partlyObservedNetwork, ...)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

nodeSampling_fit$clone(deep = FALSE)

Arguments

Observe a network partially according to a given sampling design

Description

This function draws observations in an adjacency matrix according to a given network sampling design.

Usage

observeNetwork(
  adjacencyMatrix,
  sampling,
  parameters,
  clusters = NULL,
  covariates = list(),
  similarity = l1_similarity,
  intercept = 0
)

Arguments

Details

Internal functions use future_lapply, so set your plan to 'multisession' or 'multicore' to use several cores/workers. The list of parameters control tunes more advanced features, such as the initialization, how covariates are handled in the model, and the variational EM algorithm:

• useCov logical. If covariates is not null, should they be used for the for the SBM inference (or just for the sampling)? Default is TRUE.

• clusterInit Initial method for clustering: either a character ("spectral") or a list with length(vBlocks) vectors, each with size ncol(adjacencyMatrix), providing a user-defined clustering. Default is "spectral". similarity An R x R -> R function to compute similarities between node covariates. Default is l1_similarity, that is, -abs(x-y). Only relevant when the covariates are node-centered (i.e. covariates is a list of size-N vectors).

• threshold V-EM algorithm stops stop when an optimization step changes the objective function or the parameters by less than threshold. Default is 1e-2.

• maxIter V-EM algorithm stops when the number of iteration exceeds maxIter. Default is 50.

• fixPointIter number of fix-point iterations in the V-E step. Default is 3.

• exploration character indicating the kind of exploration used among "forward", "backward", "both" or "none". Default is "both".

• iterates integer for the number of iterations during exploration. Only relevant when exploration is different from "none". Default is 1.

• trace logical for verbosity. Default is TRUE.

The different sampling designs are split into two families in which we find dyad-centered and node-centered samplings. See doi:10.1080/01621459.2018.1562934 for a complete description.

• Missing at Random (MAR)

  • dyad parameter = p = Prob(Dyad(i,j) is observed)

  • node parameter = p = Prob(Node i is observed)

  • covar-dyad": parameter = beta in R^M, such that Prob(Dyad (i,j) is observed) = logistic(parameter' covarArray (i,j, .))

  • covar-node": parameter = nu in R^M such that Prob(Node i is observed) = logistic(parameter' covarMatrix (i,)

  • snowball": parameter = number of waves with Prob(Node i is observed in the 1st wave)

• Missing Not At Random (MNAR)

  • double-standard parameter = (p0,p1) with p0 = Prob(Dyad (i,j) is observed | the dyad is equal to 0), p1 = Prob(Dyad (i,j) is observed | the dyad is equal to 1)

  • block-node parameter = c(p(1),...,p(Q)) and p(q) = Prob(Node i is observed | node i is in cluster q)

  • block-dyad parameter = c(p(1,1),...,p(Q,Q)) and p(q,l) = Prob(Edge (i,j) is observed | node i is in cluster q and node j is in cluster l)

Value

an adjacency matrix with the same dimension as the input, yet with additional NAs.

Examples

## SBM parameters
N <- 300 # number of nodes
Q <- 3   # number of clusters
pi <- rep(1,Q)/Q     # block proportion
theta <- list(mean = diag(.45,Q) + .05 ) # connectivity matrix

## simulate an unidrected binary SBM without covariate
sbm <- sbm::sampleSimpleSBM(N, pi, theta)

## Sample network data

# some sampling design and their associated parameters
sampling_parameters <- list(
   "dyad" = .3,
   "node" = .3,
   "double-standard" = c(0.4, 0.8),
   "block-node" = c(.3, .8, .5),
   "block-dyad" = theta$mean,
   "degree" = c(.01, .01),
   "snowball" = c(2,.1)
 )

observed_networks <- list()

for (sampling in names(sampling_parameters)) {
  observed_networks[[sampling]] <-
     missSBM::observeNetwork(
       adjacencyMatrix = sbm$networkData,
       sampling        = sampling,
       parameters      = sampling_parameters[[sampling]],
       cluster         = sbm$memberships
     )
}

An R6 Class used for internal representation of a partially observed network

Description

An R6 Class used for internal representation of a partially observed network

An R6 Class used for internal representation of a partially observed network

Details

This class is not exported to the user

Active bindings

Methods

Public methods

• partlyObservedNetwork$new()

• partlyObservedNetwork$clustering()

• partlyObservedNetwork$imputation()

• partlyObservedNetwork$clone()

Method new()

constructor

Usage

partlyObservedNetwork$new(
  adjacencyMatrix,
  covariates = list(),
  similarity = l1_similarity
)

Arguments

Method clustering()

method to cluster network data with missing value

Usage

partlyObservedNetwork$clustering(
  vBlocks,
  imputation = ifelse(is.null(private$phi), "median", "average")
)

Arguments

Method imputation()

basic imputation from existing clustering

Usage

partlyObservedNetwork$imputation(type = c("median", "average", "zero"))

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

partlyObservedNetwork$clone(deep = FALSE)

Arguments

Visualization for an object missSBM_fit

Description

Plot function for the various fields of a missSBM_fit: the fitted SBM (network or connectivity), and a plot monitoring the optimization.

Usage

## S3 method for class 'missSBM_fit'
plot(
  x,
  type = c("imputed", "expected", "meso", "monitoring"),
  dimLabels = list(row = "node", col = "node"),
  ...
)

Arguments

Value

a ggplot object

Prediction of a missSBM_fit (i.e. network with imputed missing dyads)

Description

Prediction of a missSBM_fit (i.e. network with imputed missing dyads)

Usage

## S3 method for class 'missSBM_fit'
predict(object, ...)

Arguments

Value

an adjacency matrix between pairs of nodes. Missing dyads are imputed with their expected values, i.e. by there estimated probabilities of connection under the missing SBM.

Class for defining a simple dyad sampler

Description

Class for defining a simple dyad sampler

Class for defining a simple dyad sampler

Super classes

missSBM::networkSampling -> missSBM::networkSampler -> missSBM::dyadSampler -> simpleDyadSampler

Methods

Public methods

• simpleDyadSampler$new()

• simpleDyadSampler$clone()

Method new()

constructor for networkSampling

Usage

simpleDyadSampler$new(
  parameters = NA,
  nbNodes = NA,
  directed = FALSE,
  covarArray = NULL,
  intercept = 0
)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

simpleDyadSampler$clone(deep = FALSE)

Arguments

Class for defining a simple node sampler

Description

Class for defining a simple node sampler

Class for defining a simple node sampler

Super classes

missSBM::networkSampling -> missSBM::networkSampler -> missSBM::nodeSampler -> simpleNodeSampler

Methods

Public methods

• simpleNodeSampler$new()

• simpleNodeSampler$clone()

Method new()

constructor for networkSampling

Usage

simpleNodeSampler$new(
  parameters = NA,
  nbNodes = NA,
  directed = FALSE,
  covarMatrix = NULL,
  intercept = 0
)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

simpleNodeSampler$clone(deep = FALSE)

Arguments

Class for defining a snowball sampler

Description

Class for defining a snowball sampler

Class for defining a snowball sampler

Super classes

missSBM::networkSampling -> missSBM::networkSampler -> missSBM::nodeSampler -> snowballSampler

Methods

Public methods

• snowballSampler$new()

• snowballSampler$clone()

Method new()

constructor for networkSampling

Usage

snowballSampler$new(parameters = NA, adjacencyMatrix = NA, directed = FALSE)

Arguments

Method clone()

The objects of this class are cloneable with this method.

Usage

snowballSampler$clone(deep = FALSE)

Arguments

Summary method for a missSBM_fit

Description

Summary method for a missSBM_fit

Usage

## S3 method for class 'missSBM_fit'
summary(object, ...)

Arguments

Value

a basic printing output

War data set

Description

This dataset contains two networks where the nodes are countries and an edge in network "belligerent" means that the two countries have been at least once at war between years 1816 to 2007 while an edge in network "alliance" means that the two countries have had a formal alliance between years 1816 to 2012. The network belligerent have less nodes since countries which have not been at war are not considered.

Usage

war

Format

A list with 2 two igraph objects, alliance and belligerent. Each graph have three attributes: 'name' (the country name), 'power' (a score related to military power: the higher, the better) and 'trade' (a score related to the trade effort between pairs of countries).

Source

networks were extracted from https://correlatesofwar.org/

References

Sarkees, Meredith Reid and Frank Wayman (2010). Resort to War: 1816 - 2007. Washington DC: CQ Press.

Gibler, Douglas M. 2009. International military alliances, 1648-2008. CQ Press

Examples

data(war)
class(war$belligerent)
igraph::gorder(war$alliance)
igraph::gorder(war$belligerent)
igraph::edges(war$alliance)
igraph::get.graph.attribute(war$alliance)