Bayes Theorem And its Applications
First and foremost, Bayes Theorem is a statement about conditional
probabilities. "Given this, what can we say about that?" It is used most
commonly in situations where one is attempting to derive conclusions about
some theory or whatever given that one has obtained a particular set of
data. Going from data to theory is hard, but it is often straightforward
to go from theory to data. Let's say you have a theory called Life, the
Universe, and Everything. The prediction for the data is always 42. Now
you conduct an experiment and get the answer 42. Does that mean that
Life, the Universe, and Everything is correct? What if I have a theory
called 2-3-7 that predicts the answer should be a combination of powers
of those prime numbers? 42 is also possible in this theory. Bayes
Theorem is no silver bullet, but it does provide a framework for inverting
the theory-prediction combination such that one can make meaningful
statements about theory given the results of some experiment or
measurements.
Bayes Theorem is not difficult to prove. It is best done with the aid of
a Venn diagram. Such a derivation is given here
.
Bayes Theorem can be applied in a wide variety of situations. The
following is collection of topics that have interested me over the years
and which I have investigated in one way or another, in one case even
writing a paper.
-
Dark Energy Survey Year 1 Cosmology Results from Large Scale Structure and
Weak Lensing. This paper uses a Bayesian Hypothesis Test to examine
whether two sets of DES data are consistent with each other under the
hypothesis that the standard Lambda-CDM model is correct. I am one of the
(many) coauthors.
-
Where is Bill Ewasko? Bill went for a day hike in Joshua Tree
National Park in June of 2010 and has not been seen since. Three days
later, a cell tower recorded a single ping from his cell phone. In spite
of intensive searching by numerous groups and individuals, he has still
not been found. I have hiked this area as well and also found nothing.
This paper uses Bayes theorem to identify a few possible remaining places
to search. [UPDATE] He has since been found - by chance.
-
Search for the SS Central America . This was a treasure hunt. Bayes
theorem was used to identify where to search for a ship that, while
carrying a cargo of gold from mines in California, sank during a hurricane
in 1857 off the coast of Georgia. Unlike the search for Bill Ewasko, this
one was successful. I have given informal talks on the story. The most
important lesson to be drawn is as follows: whenever gold is involved, it
brings out the
worst behavior in everybody.