MCMCsampler: A simple MCMC sampler class
In dennisthemenace2/Slice: Slice Bayesian Inference Engine

Description Usage Arguments Details Value See Also Examples

This is a simple MCMC sampler class that can be used to generate samples from a model.

1	sampler = new("MCMCsampler",model)

model

Constructor takes the model object create by the lexer

This sampler inherits from the base class sampler. The function takeSample can be used to sample from the model. Before the sampler starts, it creates a list with all nodes to sample from. This is done to speed up the sampling process. It then calls the sample function, with a list of likelihood objects and prior prior object. This this can be overwritten for convenience by other samplers. The takeSample function can take up to 3 parameters (nSamples,initialValues = list(),addtionalNodes=list()). The first one is how many samples are to be taken, the second one is list with names and initial values to set, the last parameter is a list of names with addtional nodes to track. This can be useful if model variables (helperNodes) are to be tracked as well.

Returns a valid sampler object.

Package Overview: Slice-Package

Base sampler class: Sampler

Slice sampler class: SliceSampler

Maximum likelihood based sampler class: LaplaceApproximation

require('Slice')

#define a regression model 
my_model_str= "
model regModel{
  #this is a comment
  mu = x*beta1 +beta0
  y~ dnorm(mu,sigma)
  sigma~dgamma(0.01,0.01)
  beta0 ~ dnorm(0,s)
  s ~ dunif(0,1)
  beta1 ~ dnorm(0,1) #one more comment
}
"

#generate some data

set.seed(1234)
n <- 100
beta   <- c(1.4,-0.8)
sigma2 <- 1
X = cbind(rnorm(n,2,1),1)
val = X 
y = matrix( val,ncol = 1)
x = matrix(X[,1],ncol = 1)
data_list = list('y'=y,'x'=x)

lex = new('Lexer') ##create lexer obj
lex$setModelString(my_model_str) #set model
lex$setModelData(data_list) # set data

lex$lexModel() ##lexx the syntax 
 
root_plate = lex$parseModel() ##create the model

mcmcSample = new('MCMCsampler',root_plate)

samplesFromProblem = mcmcSample$takeSample(500, # take 500 samples from the posterior
                                           addtionalNodes=c('mu')) # also track mu


colMeans(samplesFromProblem)
plot(samplesFromProblem[,1])