Question: 19. Consider the following two complete conditional distributions, originally analyzed by Casella and George (1992): f(x|y) yeyx, 0
19. Consider the following two complete conditional distributions, originally analyzed by Casella and George (1992):
f(x|y) ∝ ye−yx, 0 f(y|x) ∝ xe−xy, 0 (a) Obtain an estimate of the marginal distribution of X when B = 10 using the Gibbs sampler. (b) Now suppose B = ∞, so that the complete conditional distributions are ordinary (untruncated) exponential distributions. Show analytically that fx(t)=1/t is a solution to the integral equation fx(x) = fx|y(x|y)fy|t(y|t)dy fx(t)dt in this case. Would a Gibbs sampler converge to this solution? Why or why not?
Step by Step Solution
There are 3 Steps involved in it
Get step-by-step solutions from verified subject matter experts
