On sampling the degree-of-freedom of Student's-\(t\) disturbances (Q5937063): Difference between revisions

In a Bayesian analysis of a model with Student's-t disturbances developed by \textit{J. Geweke} [J. Appl. Econom. 8, S19--S40 (1993)], and \textit{C. Fernández} and \textit{M. F. J. Steel} [J. Am. Stat. Assoc. 93, No. 441, 359--371 (1998; Zbl 0910.62024)], the degree-of-freedom of Student's-t disturbances, if unknown, must be sampled from its conditional distribution. This article presents a new method for this sampling using a Metropolis-Hastings acceptance-rejection algorithm proposed by \textit{L. Tierney} [Ann. Stat. 22, No. 4, 1701--1762 (1994; Zbl 0829.62080 )]. The acceptance probabilities in both the acceptance-rejection and Metropolis-Hastings parts of this method are shown to exceed 95\%.