automultinomial (version 2.0.0)

MPLE: Maximum pseudolikelihood estimation

Description

Fits an autologistic model or automultinomial model. Takes as arguments a design matrix X, a response vector y (in factor form), and a square symmetric adjacency matrix encoding the neighborhood structure. When the number of levels of the response y is >2, the function fits a multicategory generalization of the autologistic model. For a full description of the models the package fits and a user guide, please see the vignette.

Usage

MPLE(X, y, A, ciLevel = 0.95, method = "asymptotic", burnIn = 300,
  nBoot = 500)

Arguments

X

the n-by-p design matrix

y

the response vector (required to be a factor)

A

the square symmetric adjacency matrix A encoding the neighborhood structure

ciLevel

the confidence level to be used for inference. Defaults to 0.95 for 95 percent intervals.

method

"boot" for parametric bootstrap and "asymptotic" for asymptotic confidence intervals.

burnIn

the number of burnin samples to use for the Gibbs sampler when method="boot"

nBoot

the number of bootstrap samples to use when method="boot"

Value

a fitted auto- model MPLE object

Examples

# NOT RUN {
##########generating coefficient values and data
A=igraph::get.adjacency(igraph::make_lattice(c(40,40))) #adjacency matrix A
X=cbind(rep(1,1600),matrix(rnorm(1600*4),ncol=4)) #design matrix
gamma=0.6 #correlation parameter
beta=matrix(rnorm(5)*0.3,ncol=1) #covariate parameters
y=drawSamples(beta,gamma,X,A,burnIn=10,nSamples=1)

##########fitting model
fit=MPLE(X = X,y=factor(y),A = A,ciLevel = 0.99,method = "asymptotic")

# }