Exact Posterior Simulation from the Linear Lasso regression - Université de Rennes Accéder directement au contenu
Communication Dans Un Congrès Année : 2018

Exact Posterior Simulation from the Linear Lasso regression

Résumé

The current popular method for approximate simulation from the posterior distribution of the linear Bayesian LASSO is a Gibbs sampler. It is well-known that the output analysis of an MCMC sampler is difficult due to the complex dependence amongst the states of the underlying Markov chain. Practitioners can usually only assess the convergence of MCMC samplers using heuristics. In this paper we construct a method that yields an independent and identically distributed (iid) draws from the LASSO posterior. The advantage of such exact sampling over the MCMC sampling is that there are no difficulties with the output analysis of the exact sampler, because all the simulated states are independent. The proposed sampler works well when the dimension of the parameter space is not too large, and when it is too large to permit exact sampling, the proposed method can still be used to construct an approximate MCMC sampler.
Fichier non déposé

Dates et versions

hal-01949938 , version 1 (10-12-2018)

Identifiants

  • HAL Id : hal-01949938 , version 1

Citer

Zdravko I. Botev, Yi-Lung Chen, Pierre L'Ecuyer, Shev Macnamara. Exact Posterior Simulation from the Linear Lasso regression. 2018 Winter Simulation Conference, Dec 2018, Goteborg, Sweden. ⟨hal-01949938⟩
61 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More