{
  "id": "1403.5207",
  "version": "v3",
  "published": "2014-03-20T17:21:11.000Z",
  "updated": "2015-12-08T17:22:18.000Z",
  "title": "Transdimensional Transformation based Markov Chain Monte Carlo: with Mixture Illustrations",
  "authors": [
    "Moumita Das",
    "Sourabh Bhattacharya"
  ],
  "comment": "As to be expected, and as shown in this revised version, TTMCMC completely outperforms RJMCMC in terms of mixing, acceptance rate, computational speed and automation. Also, we introduced schematic diagrams to facilitate understanding of TTMCMC, and to particularly show that \"dimension-matching\" is not necesary for TTMCMC",
  "categories": [
    "stat.CO"
  ],
  "abstract": "In this article, we propose a novel and general dimension-hopping MCMC methodology that can update all the parameters as well as the number of parameters simultaneously using simple deterministic transformations of some low-dimensional (often one-dimensional) random variable. This methodology, which has been inspired by the recent Transformation based MCMC (TMCMC) (Dutta & Bhattacharya (2014)) for updating all the parameters simultaneously in general fixed-dimensional set-ups using low-dimensional (usually one-dimensional) random variables, facilitates great speed in terms of computation time and provides high acceptance rates, thanks to the low-dimensional random variables which effectively reduce the dimension dramatically. Quite importantly, our transformation based approach provides a natural way to automate the move-types in the variable dimensional problems. We refer to this methodology as Transdimensional Transformation based Markov Chain Monte Carlo (TTMCMC). We develop the theory of TTMCMC, illustrating it with normal mixtures with unknown number of components applied to three well-known real data sets. Comparisons with RJMCMC demonstrates far superior performance of TTMCMC in terms of mixing, acceptance rate, computational speed and automation. As byproducts of our effort on the development of TTMCMC, we propose a novel methodology to summarize the posterior distributions of the mixture densities, providing a way to obtain the mode of the posterior distribution of the densities and the associated highest posterior density credible regions. Based on our method we also propose a criterion to assess convergence of variable-dimensional algorithms. These methods of summarization and convergence assessment are applicable to general problems, not just to mixtures.",
  "revisions": [
    {
      "version": "v2",
      "updated": "2014-07-29T16:50:58.000Z",
      "title": "Transdimensional Transformation based Markov Chain Monte Carlo",
      "abstract": "In this article, we propose a novel and general dimension-hopping MCMC methodology that can update all the parameters as well as the number of parameters simultaneously using simple deterministic transformations of some low-dimensional (often one-dimensional) random variable. This methodology, which has been inspired by the recent Transformation based MCMC (TMCMC) (Dutta & Bhattacharya (2014)) for updating all the parameters simultaneously in general fixed-dimensional set-ups using low-dimensional (usually one-dimensional) random variables, facilitates great speed in terms of computation time and provides high acceptance rates, thanks to the low-dimensional random variables which effectively reduce the dimension dramatically. Quite importantly, our transformation based approach provides a natural way to automate the move-types in the variable dimensional problems. We refer to this methodology as Transdimensional Transformation based Markov Chain Monte Carlo (TTMCMC). We develop the theory of TTMCMC, illustrating it with normal mixtures with unknown number of components applied to three well-known real data sets. Moreover, in the context of density estimation, we propose a novel methodology to summarize the posterior distributions of the mixture densities, providing a way to obtain the mode of the posterior distribution of the densities and the associated highest posterior density credible regions. Based on our method we also propose a criterion to assess convergence of TTMCMC. These methods of summarization and convergence assessment are generally applicable to problems involving unknown functions such as nonparametric regression and functional data analysis.",
      "comment": null,
      "journal": null,
      "doi": null
    },
    {
      "version": "v3",
      "updated": "2015-12-08T17:22:18.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "markov chain monte carlo",
      "transdimensional transformation",
      "posterior density credible regions",
      "highest posterior density credible",
      "well-known real data sets"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable",
      "adsabs": "2014arXiv1403.5207D"
    }
  }
}