ibbs sampler produces a sample that converges in distribution to a target
multidimensional random variable specified by a distribution known up to a
normalization constant. Bayesian analysis
(
Bayesian section
) is the main
field of application.
Suppose we are given a distribution
.
The variable
is multidimensional:
.
The first element
of
a sample is drawn from some distribution
.
We proceed to generate the subsequent draws
,
as
follows:
Set
.
For every
do
-
form a vector
;
-
generate a draw
from
the single dimensional conditional distribution
;
-
store the result in the new
.
Set
.
See the
[Gelman]
for the full treatment of the
subject.
|
