Statnet and latentnet

From InterSciWiki

Jump to: navigation, search


The R software package latentnet: software to fit and evaluate latent position and cluster models for statistical networks.

Fit a Latent Space Random Graph Model

The R software package statnet: software tools for the representation, visualization, analysis and simulation of social network data.

Online Users Guide: Manuals - all to appear in -- http://www.jstatsoft.org/v23 (forthcoming v24)
  1. network: A package for managing relational data in R - Butts CT (2008).
  2. ergm: A package to fit, simulate and diagnose exponential family models for networks
  3. Using ergm : Specification of exponential-family random graph models and computational tips
  4. latentnet: a package for fitting latent cluster models for networks
  5. sna: A package for social network analysis
  6. Degreenet
  7. dynamicnetwork and rSoNIA: Prototype packages for managing and animating longitudinal network data
  8. A tutorial on statnet
install.packages("statnet")
After installation you will see 
Type help(package="statnet") to get started.
Based on "statnet" project software (http://statnetproject.org).
For license and citation information see http://statnetproject.org/attribution
or type citation("statnet").
Warning messages: (Not a problen if 2.6.1 or later)
1: package 'statnet' was built under R version 2.6.2 
2: package 'network' was built under R version 2.6.2 
3: package 'ergm' was built under R version 2.6.2 

help(package="statnet")
install.statnet 
update.statnet
library(network)

Tutorial - Mark S. Handcock

Contents

[edit] ergm

For "A simple model for complex networks with arbitrary degree distribution and clustering", Mark S. Handcock and Martina Morris, see ergm in User's Guide for "ergm: A Package to Fit, Simulate and Diagnose Exponential-Family Models for Networks" David R. Hunter, Mark S. Handcock, Carter T. Butts

You may install erge directly (from the ergm article 2008) 

install.packages("ergm")
library(ergm)

For sample datasets see 

help(samplike)       #Sampson Monastery data
? faux.mesa.high     #students in grades 7-12

You can experiment with 

data(sampson)
?samplike # ?sampson also works here
summary(samplike)

[edit] Examples of ERGMs

[edit] Bernoulli and Erd˝os-R´enyi models

model1 <- ergm(samplike ~ edges)  
model1$coef   #equal logit (88/306) = −.9072.
  • the ergm command requires the formula format in R, much like other regression-like functions such as lm for linear regression or glm for generalized linear models.
network object ~ <model term 1> + <model term 2> + · · · ,

[edit] P1 Model

 model2 <- ergm(samplike ~ edges + sender + receiver + mutual, control=control.ergm(check.degeneracy=FALSE), verbose=TRUE)
summary(model2)

[edit] exogeneous covariates and dyadic independence

this model contains terms for the overall number of edges, a differential homophily effect for grade, and a main effect for sex. We may fit this model using the faux.mesa.high dataset as follows: 

data(faux.mesa.high)
model3 <- ergm(faux.mesa.high ~ edges + nodematch("Grade", diff=T) + nodefactor("Sex"))
summary(model3)

[edit] Dyadic dependence models

degenerate - see refs

[edit] Curved exponential-family models

model4 <- ergm(faux.mesa.high ~ edges + nodematch("Grade") + gwesp(0.5, fixed=TRUE), verbose=TRUE, control=control.ergm(check.degeneracy=FALSE))
summary(model4)

[edit] Statistical inference for ERGMs

[edit] Approximating an MLE

Through simulation - see

[edit] Pseudolikelihood

Retrieved from "http://intersci.ss.uci.edu/wiki/index.php/Statnet_and_latentnet"
Personal tools