BAYESIAN WITH FULL CONDITIONAL POSTERIOR DISTRIBUTION APPROACH FOR SOLUTION OF COMPLEX MODELS
Pudji Ismartini · ePrints - UNY (Yogyakarta State University) · 2014
Complex models are often arise in social field applications. Model complexity might be caused by the number of variables in the model, or due to the complex data structures. Problems are often encountered in complex models is a difficulty to obtain a solution in the model parameter estimation process. Bayesian methods can overcome these problems with the modern approach of Bayesian analysis using special simulation procedure based on the posterior distribution of parameters, i.e. the Markov Chain Monte Carlo (MCMC). Implementation of MCMC methods for Bayesian analysis requires proper sampling algorithm in order to obtain a sample from a distribution. The algorithm which is efficient and often used by MCMC is Gibbs Sampling. One of the advantages of Gibbs sampling is the generation of random variables is done using the concept of one-dimensional distribution which are structured as a form of full conditionals, i.e. the full conditional posterior distribution of parameter. This paper propose to describe the process of parameter estimation for complex model using Bayesian through full conditional posterior distribution of parameters.