Econometrics with Octave

Econometricians sweat linear algebra. Be it for linear or non-linear problems of estimation or inference, matrix algebra is a natural way of expressing these problems on paper. However, when it comes to writing computer programs to either implement tried and tested econometric procedures, or to research and prototype new routines, programming languages such as C or Fortran are more of a burden than an aid. Having to enhance the language by supplying functions for even the most primitive operations such as transposing a matrix adds extra programming effort, introduces new points of failure, and moves the level of abstraction further away from the elegant mathematical expressions. As Eddelbüttel (1996) argues, object-oriented programming provides ‘a step up’ from Fortran or C by enabling the programmer to seamlessly add new data types such as matrices, along with operations on these new data types, to the language. But with Moore’s Law still being validated by ever and ever faster processors, and, hence, ever increasing computational power, the prime reason for using compiled code, i.e. speed, becomes less relevant. Hence the growing popularity of interpreted programming languages, both, in general, as witnessed by the surge in popularity of the generalpurpose programming languages Perl and Python and, in particular, for numerical applications with strong emphasis on matrix calculus where languages such as Gauss, Matlab, Ox, R, Splus, which were reviewed by Cribari-Neto and Jensen (1997), Cribari-Neto (1997) and Cribari-Neto and Zarkos (1999), have become popular. This article introduces another interpreted language focussed on numerical applications: Octave.