• C++ Programming for Financial Engineering
    Highly recommended by thousands of MFE students. Covers essential C++ topics with applications to financial engineering. Learn more Join!
    Python for Finance with Intro to Data Science
    Gain practical understanding of Python to read, understand, and write professional Python code for your first day on the job. Learn more Join!
    An Intuition-Based Options Primer for FE
    Ideal for entry level positions interviews and graduate studies, specializing in options trading arbitrage and options valuation models. Learn more Join!

Markov Switching Multifractal - Maximum Likelihood Estimation

Joined
1/17/12
Messages
1
Points
11
Hello there,
I am currently trying to implement the Markov Switching Model from Calvet and Fisher (2004). I read some of their papers and think that this model is quite interesting to investigate it a little further. However, the estimation of parameters via maximum likelihood seems a little too complex for me.

I want to code this procedure in C later, but at first i really want to completely understand the methods they used. Therefore I tried to code the MLE in an Excel Spreadsheet. Unfortunately I made some mistakes, as I am getting weird parameter values when I try to maximize the likelihood function via Solver.

I attached the spreasheet I developed and I would be really grateful if anyone who is interested in this topic could have a look at it. The basic outline of the model can be found in the following paper up to page 57: MSM - Calvet/Fisher Paper

A brief summary of the formulas can also be found in good old Wikipedia: MSM - Wikipedia

As the model estimation grows exponentially with the volatility components k, i limited this number to 3 (again, my first step is to understand the model) and limitations for the other parameters are also specified in the attachment. The parameters that should be estimated are marked with a grey background as is the sum of the likelihood function. My supposition is that my transition matrix is incorrect (J15:Q22) and also the initial vector Pi0 (W28:AD28). The implementation is their binomial model - therefore m0 only can have two values (m0 and 2-m0).

I hope that someone will have a look at it and I think its definitely worth to give this multifractal model some attention. Thanks in advance to all those who are interested in this topic. Hopefully this is the start of a fruitful and ongoing discussion.
 

Attachments

  • Markov Switching Multifractal - MLE.xls
    464.5 KB · Views: 92
Back
Top