meleangelo / netnew

Replication package for Mele (2015)

netnew: a replication package for Mele (2015)

This is an R package for replication of the methods and analysis in A Structural Model of Segregation in Social Networks.

Installation

To install the package from the Github repository, you need to have devtools already installed.
If you don't have the devtools, install it using the following command

install.packages("devtools")

Then open devtools and use the install_github command to download and install netnew:

library(devtools)
install_github("meleangelo/netnew")

This should install the package. This works well in Windows, using RStudio.
To use the package just open it as an R library

library(netnew)

Now you should be able to run the codes.

Simulating a network model

We want to simulate a model with $n=100$ players and only direct links and indirect links utility, starting the simulations at the empty network. We want to use a local chain and get a sapmle of $R=10000$ networks obtained sampling every 15 iterations of the sampler.

library(netnew)
set.seed(1977)
n <- 100 # size of network
a <- c(-3,0,3/n) # parameters for simulations

g <- matrix(data = rbinom(n*n,1,0), nrow = n, ncol = n)
diag(g)<-0
net <- simulatemodel(model = "3params", g, a, skip = 15, samplesize = 10000)

The function simulatemodel is used to simulate some specific model. 3params has 3 parameters: direct utility, reciprocity, and indirect links. Here we set the parameter for reciprocity to zero. The result is the sample of simulated network statistics

head(net)
  links mutual indirect
1     0      0        0
2     1      0        0
3     1      0        0
4     2      0        0
5     4      0        0
6     4      0        0
tail(net)
      links mutual indirect
9995    722     26     5300
9996    721     26     5282
9997    720     26     5266
9998    720     26     5261
9999    718     26     5227
10000   719     26     5244

We can visualize the convergence of the algorithm using

plot.ts(net)

A tutorial for simulations of models

Here is a tutorial vignette that explains some more detail
vignette