On Inference of Finite Mixture of Rayleigh Distribution by Gibbs Sampler and Metropolis-Hastings

Inferential methods of statistical distributions have reached a high level of interest in recent years. However, in real life, data can follow more than one distribution, and then mixture models must be fitted to such data. One of which is a finite mixture of Rayleigh distribution that is widely used in modelling lifetime data in many fields, such as medicine, agriculture and engineering. In this paper, we proposed a new Bayesian frameworks by assuming conjugate priors for the square of the component parameters. We used this prior distribution in the classical Bayesian, Metropolis-hasting (MH) and Gibbs sampler methods. The performance of these techniques were assessed by conducting data which was generated from two and three-component mixture of the Rayleigh distribution according to several scenarios and comparing the results of the scenarios by calculating the mean of classification successful rate (MCSR) and the mean of mean square error(MMSE). The results showed that Gibbs sampler algorithm yields a better computation results than the others in terms of MMSE and MCSR.