next up previous contents
Next: Growth Model for with Up: Calculations Previous: Multiplicative Variance Growth Model

 

Acceptance Probabilities

In the context of Metropolis Hastings, in general, an acceptance probability needs to be calculated for the proposed state. Let the target distribution, the posterior, be denoted tex2html_wrap_inline3383 and the prior distribution denoted tex2html_wrap_inline2449 . Let the parameter or parameters of interest be denoted tex2html_wrap_inline2033 , with chain samples of tex2html_wrap_inline2033 denoted x. A proposed new state will be denoted tex2html_wrap_inline3393 , and the current state denoted tex2html_wrap_inline2571 . The proposal density is just tex2html_wrap_inline3397 . The data, as before, may be denoted tex2html_wrap_inline2855 , which consist of lengths for I cracks at times tex2html_wrap_inline3403 .

It is known that the acceptance probability is:

equation1579

Using Bayes, the form of the posterior is obtainable from the likelihood and prior as:

displaymath3405

and the constant of proportionality is

displaymath3407

which is clearly independent of x. Since this is the case, the ratio of the posteriors is the ratio of the products of the priors and likelihoods.





Cathal Walsh
Sat Jan 22 17:09:53 GMT 2000