LaplaceApproximation: A simple class the maximum likelihood optimizes each...
In dennisthemenace2/Slice: Slice Bayesian Inference Engine

Description Usage Arguments Details Value See Also Examples

This is a simple slice sampler that uses stepping out for interval estimation.

1	sampler = new("LaplaceApproximation",model)

model

Constructor takes the model object create by the lexer

This sampler inherits from MCMCsampler, so that it can benefit from the there implement functionality. It overwrites the sample function that gets called with the likelihood and prior. It used the R optim function to optimize the posterior and sets this value

Returns a valid sampler object.

Package Overview: Slice-Package

Base sampler class: Sampler

Slice sampler class: MCMCsampler

require(Slice)


set.seed(1234)
n <- 100
lambda = 1.125
beta   <- c(1.4,-0.5)
sigma2 <- 0.1
n_time = 10

beta_new = beta[1]
beta_new = cbind(beta_new, beta[2])

X<-cbind(matrix(rnorm(n*n_time,2,1) ,ncol = n_time) ,1)

val = X
y<- matrix( val + rnorm(n, 0,sigma2) ,ncol = 1)

colnames(y) = 'y'
x<- matrix(X[,1:(ncol(X)-1) ],ncol = n_time)
colnames(x)= paste('x_',1:n_time,sep = '')

model_str= "
model regModel{

  y ~ dnorm( mu , sigma) 
  sequ =  lambda * seq(0,weights-1) 
  mu = x 

  lambda~dgamma(0.001,0.001)
  sigma~dgamma(0.001,0.001)
  beta0 ~ dnorm(0,s)
  beta1 ~ dnorm(0,s)
  s~dgamma(0.001,0.001)
}
"

data_list = list( 'y'=y , 'x'=x ,'weights'=n_time)

lex = new('Lexer')
lex$setModelString(model_str)
lex$setModelData(data_list)

lex$lexModel()

root_plate = lex$parseModel()

laplace = new('LaplaceApproximation',root_plate)
laplaceApprox = laplace$takeSample(50)

plot(laplaceApprox[,1])