automultinomial (version 1.0.0)

simulateData: Simulate data

Description

Simulates data from the centered or uncentered binomial or multinomial distribution

Usage

simulateData(theta, X, m, n, boundary = -1, centered = FALSE, k = 2,
  burn = 40, draws = 1)

Arguments

theta

a vector of coefficients for binary response, or a matrix of coefficients for multicategory response

X

n by p design matrix

m

width of the grid

n

height of the grid

boundary

(uncentered only) -1 for no boundary conditions, i>=0 to surround grid with class i observations

centered

logical, generates data from centered model if TRUE and uncentered if FALSE

k

number of response categories

burn

number of burn-in iterations for the Gibbs sampler

draws

number of simulated samples to return

Value

an nxk matrix of responses for draws=1, and a length draws list of nxk response matrices for draws>1

Examples

# NOT RUN {
#generate outcomes on a grid using example coefficient values and design matrix
set.seed(42)
#for a multinomial response, beta and eta will have more than one column
#for a binary response, beta and eta will only have one column
#here, we the response variable takes 3 possible values, so beta and eta have two columns
beta=cbind(c(0,1.25,-1,2,-0.8),c(-3,0.25,-5,1,-0.4))

#setting the eta coefficients
eta=cbind(c(0.6,0),c(0,0.6))

#these are the final coefficient values we'll use to generate the data
theta=rbind(beta,eta)

#X matrix with 5 predictors and 900 observations for a 10x10 grid
X=cbind(rep(1,900),replicate(4,rnorm(900)))

#generate data for 30x30 grid
z=simulateData(theta,X,30,30,centered=TRUE,k=3,burn=1)
# }