hmflatlogit {bayess}

Metropolis-Hastings for the logit model under a flat prior

Description

Under the assumption that the posterior distribution is well-defined, this Metropolis-Hastings algorithm produces a sample from the posterior distribution on the logit model coefficient \beta under a flat prior.

Usage

hmflatlogit(niter, y, X, scale)

Arguments

niter

number of iterations

y

binary response variable

X

matrix of covariates with the same number of rows as y

scale

scale of the Metropolis-Hastings random walk

Value

The function produces a sample of \beta's as a matrix of size niter x p, where p is the number of covariates.

See Also

hmflatprobit

Examples

data(bank)
bank=as.matrix(bank)
y=bank[,5]
X=bank[,1:4]
flatlogit=hmflatlogit(1000,y,X,1)
par(mfrow=c(1,3),mar=1+c(1.5,1.5,1.5,1.5))
plot(flatlogit[,1],type="l",xlab="Iterations",ylab=expression(beta[1]))
hist(flatlogit[101:1000,1],nclass=50,prob=TRUE,main="",xlab=expression(beta[1]))
acf(flatlogit[101:1000,1],lag=10,main="",ylab="Autocorrelation",ci=FALSE)