Y by
Suppose we are given i.i.d.\ observations
from a distribution with probability density function (PDF)
for
, where the parameter
has a prior distribution with PDF
. Consider the following two approaches to Bayesian updating:
(1) Let
be the complete data vector. Denote the posterior PDF as
, where
, obtained by applying Bayes' rule to the full dataset at once.
(2) Start with prior
. For each
, let
be the current prior and update it using observation
to obtain the new posterior:

Are the final posteriors
from part (a) and
from part (b) the same? Provide a proof or a counterexample.
Here is the proof I've written:
Do you guys think this is rigorous enough? What would you change?





(1) Let



(2) Start with prior





Are the final posteriors


Here is the proof I've written:
Do you guys think this is rigorous enough? What would you change?