The parameter in the growth model has a hyperparameter
such that the
are exchangeable with the following
distribution for the parameters:
The proposal density for the is Gaussian,
with mean
and sampler variance tuned for acceptance rate. Thus it is
symmetric in
, that is to say that in this case
. Thus the acceptance probability reduces to the ratio of
products as above:
And the prior comes from the above. That is, the product becomes:
which can be simplified for computational purposes and by cancellation leaving the following:
The acceptance probability is the minimum of this and 1. As an aside, note that when
carrying out such calculations, if then
In this way, one avoids floating point errors which may occur when evaluating the acceptance
probability, as they did for this author.