Demog Blog‎ > ‎

Eddie Hunsinger's Applied Demography Toolbox

posted Jun 25, 2011, 4:28 AM by Tim Riffe   [ updated Jun 28, 2011, 5:57 AM ]
I've long been aware of some CRAN-posted packages for use in demography: "demography", by Rob Hyndman (and his many other packages posted at his personal page), "demogR", authored and maintained by various of the good folks in Berkeley, "MortalitySmooth" by giancarlo, and a few packages used by people from various disciplines but that demographers also use heavily, like "survival" and "epi". I did not know of a decent repository for the many random scripts and spreadsheets that random demographers use and post randomly online. Behold! Such a site has been founded, and it is maintained by Eddie Hunsinger, Alaska State Demographer, here.

Aside 1: That's right, Alaska has at least 1 state demographer. Does this mean that there's hope for me in Michigan after all? I mean, demographers ought to scale by population right? I've always been envious of my European colleagues from small countries or autonomous communities that have govt jobs for demographers- it's an automatic in! I would totally dig the Michigan govt demographer job. I mean, seriously, Michigan has population issues!

Aside 2: Did you notice that he hosts it on his old Berkeley webspace? I used to have one like that at the University of Michigan, but they erased it after graduation, and I lost tons of papers and class projects that were posted there, not that I would have ever gone back and looked at them-- the point is, wow, Berkeley is less stingy than Michigan! C'mon umich!

There appear to be tons of rock'n R scripts and Excel sheets and SAS macros and whatnot just waiting to be explored. Most folks write one-use scripts to solve little problems that indeed are recurring in the discipline, but that we deem not worth packaging, either due to lack of confidence, or because they appear too narrow, or we don't feel like cleaning them up, or because writing package documentation is a pain, and making the blind leap into Rcmd build --binary packagename is going way too far. These scripts just don't make it online often enough. The same goes for the syntax used to reproduce studies: I see lots of PAA presentations of studies done in R, and they all have their code posted hither and tither. Hellz, my site is off in hither-tither land...

 If GoogleSites would let me, I'd post some online with the .R suffix, then they could be called directly from R using source(url("http...")). Eddie's page has this feature, i.e. that of posting the scripts directly with the .R suffix, so they can be called in this way. Maybe another good idea would be to put up html documentation and examples, per the normal R guidelines, and eventually move toward packaging them? The page is of course not R-only, and that's really good for attracting people in. Another idealistic suggestion: Rosetta-stone-type syntax translation of things that do get submitted- i.e. how to do this or that spreadsheet in R or SAS and vice versa. Maybe that's going too far? Wikify it? Anyway, it appears to be a newish site, with lots of goodies, go check it out and contribute.

Here's the link again: