Introduction

In principle, however, it is difficult to publish a paper that describes an original way to make a plot, to print output, to perform a complicated analysis in a particular package. Some of it is published in the American Statistician or the Proceedings of the Interface or Compstat; otherwise it winds up in conference proceedings, newsletters or electronic mailing lists devoted to a single package. If there is too much emphasis on the actual program, it cannot go into the usual statistics journals, not even into the computational statistics journals. This is quite appropriate, but it is also unfortunate that there is no outlet at all for such papers (or manuals, if you like).

The Journal of Statistical Software is

• electronic and freely available,
• done in HTML (using Postscript or GIF where necessary),
• interactive, where possible,
• peer reviewed.

The Information for Authors section of JCGS contains this line: "Articles describing or comparing software may be appropriate if they introduce new concepts or algorithms, or otherwise improve existing knowledge about the software." This line, a bit expanded, would be appropriate for JSS as well. There may be no way to avoid this kind of overlap, except by emphasizing that JSS is a forum specifically for articles about software, whereas it's a sideline for other journals. If JSS gets a wide enough readership, authors may eventually see an advantage in a hands-on outlet for publication. What distinguishes an article in JSS from a submission to Statlib? Here's one possible answer. A port of statistical code from Fortran to Xlisp, by itself, for example, would just go to Statlib. If the author can demonstrate that the addition of the code to an Xlisp-Stat environment allows a new or better way of tackling some task, however, this would make the work interesting to JSS. JSS is a popular idea, I think, partly or mainly because of its emphasis on statistical practice. To be worthwhile, a piece of work should show how it makes a contribution to improving statistical practice. The example articles pretty much do this: they present a problem and give a solution, contrast it with existing solutions, and explain how and why the code is useful. The wider the applicability of the code (i.e., the greater the potential of the code to affect the reader's work), the more interesting the article.