I'm trying to run an FLS regression on a dataset and I basically want to constrain it so that at any given time period, the sum of the coefficients/betas adds up to 1 as a condition.
Constraints in regression models in SHAZAM are possible using the RESTRICT Command however constraining an FLS is not supported by this command.
There are some recent papers that describe how this might be possible. Please see the following important page related to FLS: http://www2.econ.iastate.edu/tesfatsi/flshome.htm. One interesting reference on that page is:
Michael Markov, Vadim Mottl, and Ilya Muchnik (2004), "Principles of Nonstationary Regression Estimation: A New Approach to Dynamic Multi-Factor Models in Finance"
Thanks for the help. However, I actually got to SHAZAM after exploring this paper in detail. That's the exact process/model that I want to replicate on SHAZAM, the only small missing piece is that the asset weights have to add up to 1 since they're part of a whole. Do you know of a workaround?
For a fixed constraint you can substitute the constraint into the equation. e.g.
Y = a + bX + (1-b)Z
becomes
Y = (a-1) + b(X-Z)
It might be worth trying to contact the authors and see if they will share their SHAZAM code.
Asked: 2014-01-09 10:00:52 +0000
Seen: 1,167 times
Last updated: Jan 24 '14
Copyright SHAZAM Analytics, 2018 Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.