We shouldn’t’ve called it “Stan”; I should’ve listened to Bob and Hadley

Hadley told me that one reason he came up with the name ggplot was that it would be uniquely findable on Google. When we were writing Stan and I suggested naming it Stan, Bob pointed out the googling argument but I just loved the name Stan, I loved the Ulam connection and having this friendly name, I insisted and Bob and others were like, ok, that’s a bad idea but whatever. In retrospect, though, I was wrong, both to want Stan as a name and not to listen more carefully to the others. We could’ve given it a random name of the xkcd variety, or called it something like Bayes Engine or BDAcalc or whatever. Too late to change it now, though!

P.S. Alan Edelman made the same mistake with Julia. And Python and Java too. I guess the people who named those programs might argue that it’s fine to use a common word as a name. And maybe it is. Stan’s a success anyway. But, right now, as I’m searching online for Stan successes and googling Stan + Nuts, Stan + Bayes, etc., I’m wishing we’d chosen a unique character string.

42 thoughts on “We shouldn’t’ve called it “Stan”; I should’ve listened to Bob and Hadley

  1. When I heard high schoolers talking about Stan, I thought Bayesian statistics was really catching on with zoomers. But it turns out that it’s just a hip term from an Eminem song that means “obsessive fan.”

  2. When I search for “R” stuff I usually add “CRAN” and it seems to work. So perhaps you just need a more unique term associated strongly enough with Stan that people can add.

  3. Andrew, Hadley, and I were in Andrew’s office discussing the name before the first Stan release, so presumably Matt, Daniel, Jiqiang, and Ben were there, too—we used to have big meetings in Andrew’s office.

    Searching on Google Scholar, I tend to use something like [“mc-stan.org” OR “stan development team”]. The quotes make sure you get exactly that string in that order, and the OR is what it says it is. The term “brms” is too short—lots of false positive hits on the acronym. But “rstanarm” should be OK. If you want NUTS citations in addition to Stan citation, then “no-u-turn sampler” is unlikely to produce false positives.

    Maybe you thought we’d be more popular. Java and Python won their naming battles. All of the top hits for me in a fresh browser are for the languages. Julia’s close—it has 9/10 top hits. Sadly, 9 of 10 hits for Stan are for the song. It didn’t help that Merriam-Webster added the Eminem sense of “Stan” to the dictionary. That pushed us back down the search results compared to where we were a couple years ago.

    • I think Stan is okay. Maybe not the best name in the world, but at least it isn’t something interminably lame and descriptive like BDAtool. The world could use a bit more idiosyncrasy in the world.

      If Stan were only a modest success, the search issue would be a problem. Personally, I have been able to find things just fine. My Stan searches almost always include a bunch of technical terms that weed out completely unrelated “Stans.”

    • Yes. Python was created in 1989, Java in the early 90s, and Google was created in 1998. Of course, there were other search engines before that, but I think it’s fair to argue that Python and Java were created well before anyone would have worried about ease of finding either via an internet search engine.

  4. On the other hand, you are giving Google too little credit. It does not naively match character strings. Long long ago, they were proud to mention that “apple” and “apples” would return very different results. Today, adding “in r” or “in stan” to a query involving other statistical terms typically shows you want you want in the first page. [I know, I know, not always]. This will only get better, so no need to change the name. Which is a great name, by the way.

    • Google tailors search results to your search habits. When I googled “latex” just now, I found that in the first 5 pages of search results, all but 4 links were related to the text processing system. Were your results different??

      • You can try to search http://www.startpage.com then. Startpage acts as a proxy between you and Google, meaning everyone’s requests are mixed in Google’s point of view. That’s the closest approximation of a naive Google I could get without erasing my browser’s history, cookies and all.

        On Startpage, googling R actually finds the R project website, same for LaTeX. LaTeX and latex yield the same results. Image Search is a different thing indeed, but I still have to find a case in which this feature would be crucial in a statistical context.

  5. What’s that old quote about naming things being one of the two hard things in software development?

    I ran into a similar problem with the retrodesign package. Ideally it’d have been something involving s/m error, but…
    1. SM (server management) error is a common problem with some types of databases, so it’s hard to google alone.
    2. There’s already an sm package, so adding “R” to the query makes it worse, not better.
    3. Sign and Magnitude aren’t exactly unique words.

    So retrodesign isn’t a great name, but at least it pops up with the 1950s polka dot dresses in the first few search results.

    • If I were to rename anything with assured uniqueness, I’d lean toward upstart rock band tendencies on the subject. Certainly the vast majority of those band names (or DJs even?) remain wholly unique in the world ‘o’ search. Maybe using a relatively uncommon name with a questionably appropriate consonant blend. Maybe ‘F{B}uck’ … Any takers?

      Jest aside, I think so many go back to their original naming of something/someone and really question what they’re were thinking at the time. Second guessing the name…. weary of the original….etc etc

      Stan is fine … but StanSchmaltz *could* be more Pinterest worthy :)

  6. I usually just add words like “MCMC”, “Bayesian”, “model”, “sampling”, or “prior”, depending on what specifically I’m looking for. I guess I don’t really know what I’m not retrieving, but I usually find something useful. At one point, there was a web site with an R-specific search engine, but I can’t find it now. Didn’t use it much, I guess.

  7. 1. I know from experience that terms beginning with “z” that aren’t actual words tend to be easy to Google.
    2. It’s not that unusual for programs to change names. Wasn’t WinBugs originally Bugs? The name change occurred with a major shift to the Windows platform, IIRC. I’m also a frequent user of Alteryx, which had a different name I’ve now forgotten when it was first introduced in the early 2000s.

    So, when you make a major efficiency/speed update in Stan, why not make this zStan (for Zippy Stan)?

Leave a Reply

Your email address will not be published. Required fields are marked *