Jon Faust
mostly economics

home

 Research Programming

Science and replicability

The hallmark of science is replicability. When your results come from computer analysis, replicability can come two ways: you write an extended account of what you did on the computer or you simply provide a script that will re-create everything you did. The descriptive approach is clumsy and imprecise and sufficiently difficult that most folks don't bother. That is, they punt on the whole science thing. This page has some pointers for scripting work in Matlab, LaTeX, and other programs.

The main items here are a set of rules of thumb and then two examples. The first is a set of programs of increasing complexity computing for computing and reporting a simple set of Monte Carlo experiments. The second shows a method for automatically downloading, plotting, and reporting macro data in batch mode using gawk, wget, matlab, acrobat, and latex.

Stuff

  • 15 rules of thumb for effective research programming. »[pdf 40k]
  • An example using Matlab and LaTeX in an integrated way. Unzip and look at readme.txt. »[zip 61k]
  • Batch downloading and analysis of data using wget, matlab, LaTeX, etc. Unzip and look at readme.txt. »[zip 113k]