|
| << vorige | volgende >> |
Tijdschrift beschrijving
Alle jaargangen van het bijbehorende tijdschrift
Alle afleveringen van het bijbehorende jaargang
Alle artikelen van de bijbehorende aflevering
Details van artikel 18 van 20 gevonden artikelen
| Titel: |
On the performance of the gibbs sampler for the multivariate normal distribution |
| Auteur: |
Menzefricke, Ulrich |
| Verschenen in: |
Communications in statistics |
| Paginering: | Jaargang 24 (1995) nr. 1 pagina's 191-213 |
| Jaar: |
1995 |
| Inhoud: |
In this paper, the performance of the Gibbs sampler is examined in a setting where analytical results are readily available, the multivariate normal distribution. Two questions are studied: (i) How many initial iterates should be discarded until the Gibbs sampler converges, and (ii) if every k-th iterate is used, how large should k be so that the retained iterates are nearly independent? Formulas are reviewed for the mean vector and the covariance matrix at iteration j, as is an expression for the autocorrelations between iterates which are k iterations apart. The case when all correlations of the multivariate normal distribution are the same is studied in some detail, and an example is given. |
| Uitgever: |
Taylor & Francis |
| Bronbestand: |
Elektronische Wetenschappelijke Tijdschriften |
Details van artikel 18 van 20 gevonden artikelen
| << vorige | volgende >> |