banam: Bayesian estimation of the network autocorrelation model

In jomulder/BANAM: Bayesian Analysis of the Network Autocorrelation Model

Bayesian estimation of the network autocorrelation model

Description

The banam function can be used for Bayesian estimation of the network autocorrelation model (NAM). In the case of a single weight matrix, a flat prior, the independence Jeffreys prior, and a normal prior can be specified for the network autocorrelation parameter. In the case of multiple weight matrices, a multivariate normal prior can be specified.

Usage

banam(
  y,
  X,
  W,
  prior = "flat",
  priormean = NULL,
  priorsigma = NULL,
  postdraws = 5000,
  burnin = 1000
)

Arguments

y	A numeric vector containing the observations of the outcome variable.
X	The design matrix of the predictor variables. If absent a column of ones is automatically to model the intercept.
W	A weight matrix (in the case of a NAM with a single weight matrix) or a list of weight matrices (in the case of a NAM with multiple weight matrices).
prior	A character string specifying which prior to use in the case of a NAM with a single weight matrix. The options are 'flat', 'IJ', and 'normal', for a flat prior, the independence Jeffreys prior, and a normal prior, respectively.
priormean	A scalar (or vector) specifying the prior mean(s) of the network autocorrelation(s). In the univariate case, the default prior mean is 0.36 which is the weakly informative prior mean from Dittrich, Leenders, & Mulder (2017). In the multivariate case zero prior means are used by default.
priorsigma	A scalar (or matrix) specofying the prior variance (or prior covariance matrix) of the network autocorrelation(s). In the univariate case, the default prior variance is 0.49 which is the weakly informative prior variance from Dittrich, Leenders, & Mulder (2017). In the multivariate case the prior covariance matrix is the identity matrix by default.
postdraws	An integer specifying the number of posterior draws after burn-in.
burnin	An integer specifying the number of draws for burn-in.

Value

The output is an object of class banam. For users of BANAM, the following are the useful objects:

• rho.draws Matrix of posterior draws for the network autocorrelation parameter(s).

• beta.draws Matrix of posterior draws for the coefficients.

• sigma2.draws Matrix of posterior draws for the error variance.

References

Dittrich, D., Leenders, R.Th.A.J. Leenders, & Mulder, J. (2017). Bayesian estimation of the network autocorrelation model. Social Network, 48, 213–236. https://doi.org/10.1016/j.socnet.2016.09.002

Dittrich, D., Leenders, R.Th.A.J. Leenders, & Mulder, J. (2020). Network Autocorrelation Modeling: Bayesian Techniques for Estimating and Testing Multiple Network Autocorrelations. Sociological Methodology. https://doi.org/10.1177/0081175020913899

Examples

#example analyses
#generate example data
set.seed(3)
n <- 50
d1 <- .2
Wadj1 <- sna::rgraph(n, tprob=d1, mode="graph")
W1 <- sna::make.stochastic(Wadj1, mode="row")
d2 <- .4
Wadj2 <- sna::rgraph(n, tprob=d2, mode="graph")
W2 <- sna::make.stochastic(Wadj2, mode="row")
# set rho, beta, sigma2, and generate y
rho1 <- .3
K <- 3
beta <- rnorm(K)
sigma2 <- 1
X <- matrix(c(rep(1, n), rnorm(n*(K-1))), nrow=n, ncol=K)
y <- c((solve(diag(n) - rho1*W1))%*%(X%*%beta + rnorm(n)))

#Bayesian estimation of NAM with a single weight matrix using a flat prior for rho
best1 <- banam(y,X,W1)
print(best1)

#Bayesian estimation of NAM with two weight matrices using standard normal priors
best2 <- banam(y,X,W=list(W1,W2))
print(best2)

#Bayes factor testing of equality/order hypotheses using environment of package 'BFpack'
BFbest2 <- BF(best2,hypothesis="rho1>rho2>0; rho1=rho2>0; rho1=rho2=0")

jomulder/BANAM documentation built on July 24, 2022, 9:25 p.m.