Dear Sven,

I hope you don't get frustrated that I do not give an answer to the exact questions you posed, but I address something more fundamental.

The use of Bayesian methods has methodological advantages as such. These methods are used in such diverse fields like astrophysics, and the study of earthquakes induced by gas production (Groningen field.). Making different models and determine which is best by computing the Bayesfactors is a conceptually simple method and more direct than frequentist hypothesis testing.  This seems to sink in with the scientific community. Might that be a reason why BCPA  as a promising and potentially superior methodology features?

Hey Manuel,

thanks for the reply anyway!

You are totally right with what you say. Moreover, we can incorporate several sources of uncertainty. The calculation of probabilities for a certain model or a point in time being a change point in the case of BCPA displays these uncertainty as well. But for me it seems that in the financial practice this methodology is not much attended, for whatever reason. And apart from bayesian model selection, bayesian methods do not seem to produce better results,  e.g. in Portfolio and risk Management. See bayesian shrinkage estimators as example.

In general i was wondering why there has not been much more effort to use BCPA in finance as on the one hand there is nessecity for non-linearities and on the other hand there are several fields of research that successfully apply BCPA.

Sven

Dear Prof. Theis,

                             I have made a distinction between passive Bayes Rule and active Bayes Rule in my PhD thesis titled "Bounded Rationality and Arrow Debreu Economies" at NYU by making the updating of probability beliefs by Lagrangean Actions and Markovian Controlled Actions in Econophysics methods and implicitly String Theory Methods developed further later following on research done by Radner & Rothschild and Kihlstrom, Mirman, Daelbaen, Nyarko, Olson and so on. There are anumber of ways in which this can be done and I have shown that the Asset Pricing literature is also dependant dynamically on this element of technology. You can take a look at my most recent papers like The String Theory of Indian Stock Market Nanostructure System on my Research Gate page. I have constructed the information and string perturbation energy field (tachyon condensation energy field with genetic effort) for such system and the innovative use of Bayesisn Learning Technology methods is necessary for these derivations which are experimental;ly testable. Earl Chair Prof. Dr. SKM QC EPS Fellow (In) MES MRES MAICTE

Dear Mr. Mallick,

your research sounds very interesting. However, I would prefer to go back to the bayesian change point method and its practical use in finacne.

BCPA is able to capture non-linearities in any model. Assume a gaussian process with changes in mu and sigma. Then we can use BCPA to detect turning points in the mean and volatility clustering. However, in finance the use of the turning point algorithm of Bry/Boschan (1973) and the popular GARCH model of Engle (1982) is common.

Might this be due to the general aversion of practitioners to use bayesian methods? Or is it that the method is to complex and not intuitive (at least for frequentist statisticians)?

I would really like to shed light, not only on the specific BCPA method, but on the use of bayesian methods in practical finance in general. Especially in the light of the recent announcement of the ASA on the use of p-values.

Sven Thies

Maybe it is not that relevant, particularly so because it is an old question. My fields (earth sciences, ecology, and environment sciences) see heavy use of changepoint analysis, both Bayesian and frequentist approaches. I did have a few papers on this: in one paper (more info at https://github.com/zhaokg/Rbeast), I developed a Bayesian Changepoint algorithm algorithm called BEAST and it definitely provides alternative insights not possible from the frequentist approach. The key is not Bayesian itself but the Bayesian model averaging. My method has been used scientifically for education finance and practically for technical analysis in stock trading. I kinda understood where you came from because it took big pains for me to advocate Bayesian algorithms to many peers.