GNU Octave – Open Source Matlab Alternative

In some research I am working on with a friend at IBM Research, we’ve needed to use numerical approaches to solve quite a complex Markov chain model.  We are modeling human resource allocation for service-based organizations (e.g. consulting companies, construction companies, etc.) as a stochastic process and the model state space becomes enormous very quickly.  Looking to solve a linear system of equations to determine the steady state probabilities, we started looking into Matlab’s “\” backslash matrix division operator.  This post on Google answers actually gives a good run down on how to solve a Markov chain using this operator.  Instead of purchasing Matlab, I decided to try GNU Octave for running these calculations and have been delighted with the results.  Like the open source econometrics packages I reviewed earlier, Octave is an excellent alternative to the commercial software.   It supports most of the Matlab commands, however users should be aware of the differences.  The performance is top notch, solved a system of linear equations with a ~2000×2000 state matrix in about 20 seconds.   And it includes either GNUPlot or an alternative Java based graphics package for creating plots.  I give Octave a big thumbs up and definitely recommend it as a great open source alternative to Matlab.

Free econometrics software that’s actually worth the price!

In the last part of my dissertation I’ve been using a time series econometrics method called vector autoregression to empirically model the dynamics of innovation of information technology components, products, and infrastructure (sounds pretty awesome doesn’t it). Because I’m still a poor graduate student for the next few months and CSOM won’t let me keep any software I purchase with my research stipend – I’ve been exploring open source and free econometrics software. I found two packages that are quite good and well worth the effort to download and try out.