acceptance ratio

[Douwe Kamper]

Hey all,

It isn't clear to me what accept reject rule is used for emcee. In Goodman and Weare's (2010) paper they say the following when describing their algorithm: " accept or reject this move using the appropriate Metropolis Hastings rule ". 

However, from reading Foreman-Mackey et.al. (2013) it is not clear to me whether emcee also uses the Metropolis-Hastings accept-reject rule. The way that you have to specify parameter limits by having your posterior log-probability function returning negative infinity outside parameter bounds (see https://emcee.readthedocs.io/en/stable/user/faq/), makes me think emcee uses the regular Metropolis accept-reject rule instead. As, the sampler appears to not be directly aware of the parameter bounds set by the user. 

Kind Regards, 

Douwe