These notes are an informal, first installment in an ongoing project to develop a convenient template for computational experimentation in econometrics. The approach is illustrated by means of an example based on some current research with Steve Portnoy on improving the speed of quantile regression algorithms. The computations are carried out in SPLUS, but similar techniques could be adapted for any modern computing environment designed for statistical applications. The objective is to provide a reasonably automatic, almost painless, way to make experimental results self-documenting and reproducible. With minor modifications the same approach could be adapted to empirical applications.
August 31, 1997 . This paper was prepared using hardware and software supported by NSF Grants SBR 93-20555 and SBR 95-12440. The latter grant served to initiatate the Econometrics Lab at UIUC. Further details on the Lab and related research activities may be found at the URL http://www.econ.uiuc.edu.