mh_gibbs_2pl: mh_gibbs_2pl Function

In nmolanog/bayesuirt:

Description

this functions implements the metropolis-hastings within gibbs algorithm proposed in molano thesis.

Usage

mh_gibbs_2pl(dat, idname, psi0, theta0, B = diag(c(rep(9, dim(X)[2]), rep(1,
  dim(X)[2]))), b = c(rep(0, length(psi0))), Nc = 10000)

Arguments

dat	a data frame containing test info. rows for individuals and columns for items. aditionally, a variable which identify individuals must be supied.
idname	atomic character giving the name of the id variable in dat.
psi0	a vector of initial values for parameter psi0. See nlf_2pl documentation.
theta0	vector of initial values for latent traits. See nlf_2pl documentation.
B	Prior covariance matrix for psi.
b	Prior expected values for psi.
Nc	number of mcmc iterations.

Value

a list with the following objects:

• fixp: a matrix where each row corresponds to an mcmc simulation of psi parameters.

• theta: a matrix where each row corresponds to an mcmc simulation of latent traits.

• Dev: a vector where deviance is calculated for each mcmc iteration.

• B: Prior covariance matrix used for psi.

• b: Prior Expected values used for psi.

Examples

nitems<-7
nind<-100
sort(alpha_o<-runif(nitems,0.6,1.5))
sort(beta_o<-runif(nitems,-2.4,2.4))
d_o<- -alpha_o*beta_o

summary(theta_o<-rnorm(nind))
temp.b<-matrix(beta_o,ncol=nitems,nrow=nind, byrow =T)
temp.a<-matrix(alpha_o,ncol=nitems,nrow=nind, byrow =T)
temp.t<-matrix(theta_o,ncol=nitems,nrow=nind)
eta<-temp.a*(theta_o-temp.b)
pmod<-exp(eta)/(1+exp(eta))
test<-data.frame(ifelse(runif(nitems*nind)<pmod,1,0))
colnames(test)<-paste("t",1:nitems,sep="")
sort(apply(test,2,function(i)sum(i)*100/length(i)))
test<-data.frame(test,id=1:nind)
bres<-mh_gibbs_2pl(test,"id",psi0=c(d_o,log(alpha_o)),theta_o,Nc=1000)

nmolanog/bayesuirt documentation built on May 23, 2019, 8:52 a.m.