Creates a series of plots showing the proportion of proposed moves to accepted moves over the duration of the MCMC analysis for each parameter updated via a random-walk sampler.
plot_all_acceptance_rates(MCMC.output)
The standard MCMC output file generated from a BEDASSLE run.
Gideon Bradburd
For optimal mixing, between ~20
samplers should be accepted. If the acceptance rates fall outside that range, this
function will automatically highlight that parameter as a potential instance of poor
mixing. If the acceptance rates are too low, then for subsequent analyses the user
should decrease the scale of the tuning parameter (or "std," as in, e.g.,
aD_std), and if acceptance rates are too high, the user should increase
the scale of the tuning parameter. The scale of the tuning parameter is the standard
deviation of the normal distribution from which the small random variable is drawn
and added to the current parameter value to propose a move. If the acceptance rate
has not plateaued by the end of an analysis, it is an indication that the chain may
still be "going somewhere" in parameter space, and subsequent analyses should be
performed.