Sie können Bookmarks mittels Listen verwalten, loggen Sie sich dafür bitte in Ihr SLUB Benutzerkonto ein.
Medientyp:
E-Artikel
Titel:
An Analysis of Swendsen-Wang and Related Sampling Methods
Beteiligte:
Fishman, George S.
Erschienen:
Blackwell Publishers, 1999
Erschienen in:
Journal of the Royal Statistical Society. Series B (Statistical Methodology), 61 (1999) 3, Seite 623-641
Sprache:
Englisch
ISSN:
1369-7412;
1467-9868
Entstehung:
Anmerkungen:
Beschreibung:
<p>Convergence rates, statistical efficiency and sampling costs are studied for the original and extended Swendsen-Wang methods of generating a sample path {S<sub>j</sub>, j ⩾ 1} with equilibrium distribution π, with r distinct elements, on a finite state space X of size N<sub>1</sub>. Given S<sub>j-1</sub>, each method uses auxiliary random variables to identify the subset of X from which S<sub>j</sub> is to be randomly sampled. Let π<sub>min</sub> and π<sub>max</sub> denote respectively the smallest and largest elements in π and let N<sub>r</sub> denote the number of elements in π with value π<sub>max</sub>. For a single auxiliary variable, uniform sampling from the subset and (N<sub>1</sub> - N<sub>r</sub>)π<sub>min</sub> + N<sub>r</sub>π<sub>max</sub> ≈ 1, our results show rapid convergence and high statistical efficiency for large π<sub>min</sub>/π<sub>max</sub> or N<sub>r</sub>/N<sub>1</sub> and slow convergence and poor statistical efficiency for small π<sub>min</sub>/π<sub>max</sub> and N<sub>r</sub>/N<sub>1</sub>. Other examples provide additional insight. For extended Swendsen-Wang methods with non-uniform subset sampling, the analysis identifies the properties of a decomposition of π(x) that favour fast convergence and high statistical efficiency. In the absence of exploitable special structure, subset sampling can be costly regardless of which of these methods is employed.</p>