File:part1-fig-04-gibbs-staircase.png

Figure file · Markov chains, Gibbs sampling, and Metropolis–Hastings

One Gibbs path through a bivariate normal target with correlation 0.95. Updating X_1\mid X_2 moves horizontally and updating X_2\mid X_1 moves vertically. The narrow diagonal target therefore turns exact conditional simulation into a staircase: every update is valid, but successive states remain strongly dependent.

One Gibbs path through a bivariate normal target with correlation 0.95. Updating X_1\mid X_2 moves horizontally and updating X_2\mid X_1 moves vertically. The narrow diagonal target therefore turns exact conditional simulation into a staircase: every update is valid, but successive states remain strongly dependent.

Size of this preview: 559 × 423 pixels. Select it to open the high-resolution file.

Summary

Description. One Gibbs path through a bivariate normal target with correlation 0.95. Updating X_1\mid X_2 moves horizontally and updating X_2\mid X_1 moves vertically. The narrow diagonal target therefore turns exact conditional simulation into a staircase: every update is valid, but successive states remain strongly dependent.

Appears in. Markov chains, Gibbs sampling, and Metropolis–Hastings

Made by. A figure from the course chapter.

Download

File usage

These 2 pages show the file: