14
$\begingroup$

To be more specific, I regularly encounter acronyms which are either ambiguous (e.g. is LDA 'Linear Discriminant Analysis' or 'Latent Dirichlet Analysis'?) or rather obscure (either because of a specialized area or just an idiosyncratic use of terminology - and googling is often little help in either case).

Either ambiguity or obscurity can lead to answering a different question than intended, or significant delays in clarifying an answer.

I try to encourage people to spell things out on first use in a question (subsequent use of abbreviations is okay), but it often feels like I'm trying to close the stable door to the sound of hoofbeats fading over the hill.

On the other hand I don't want to stop people using really standard abbreviations, but I find it very hard to draw a line, which makes me lean toward almost always spelling it out.

What do people think? Is it sensible to have that in the help? Is this perhaps too minor an issue?

$\endgroup$
  • 4
    $\begingroup$ This arises frequently. I have noticed the LDA ambiguity several times. It is part of the permanent issue that too many assume that terminology known locally is also known globally. Another example is IV, which is independent or instrumental variable, requiring effort to work out the intended meaning. I support the idea of spelling out that you should spell things out, although how much difference it would make in practice is moot. (Precisely when an abbreviation becomes an acronym would need to be considered in drafting something.) $\endgroup$ – Nick Cox Aug 7 '13 at 7:39
  • 2
    $\begingroup$ DV I just seen used to mean data visualization: stats.stackexchange.com/questions/66733/… If you thought it definitely meant "dependent variable"... $\endgroup$ – Nick Cox Aug 7 '13 at 13:37
  • 2
    $\begingroup$ Here's a fun one: Does GLM stand for general linear model or generalized linear model? $\endgroup$ – gung - Reinstate Monica Aug 7 '13 at 15:06
  • 1
    $\begingroup$ @Macro In respect of GLMs ... I wish it were so, but unfortunately, I have seen the general linear model written as "GLM" in course notes --- and it is even in Wikpedia(!); I recall asking a poster of a stats question (probably here on CV) to change a question that did the same. So even with GLM it might be better for posters to consider the possibility that it might be ambiguous. $\endgroup$ – Glen_b -Reinstate Monica Aug 8 '13 at 23:08
  • $\begingroup$ @Glen_b, I thought it was clear from the parenthetical that, while I always use "GLM" to mean generalized linear model, my rebuke of gung was 100% a joke and not meant to inspire a debate. I'm not surprised it has been referred to as a GLM before but I'd never seen it. In my experience these are just called linear models. Yes, I saw the wikipedia page but frankly the page (at a glance) looks a bit sketchy so I didn't consider it authoritative. I've deleted the previous comment and am conceding this debate I never meant to start. Cheers. $\endgroup$ – Macro Aug 9 '13 at 17:25
  • $\begingroup$ My apologies, Macro, I missed that it was a joke. $\endgroup$ – Glen_b -Reinstate Monica Aug 9 '13 at 21:45
  • 2
    $\begingroup$ I agree that it is an issue. Possibly even better than "try to avoid acronyms", though would be "spell out all acronyms on first use". (A) There is likely some subset of people that will only recognize (or more readily recognize) the acronym (and maybe learn something by seeing the long form and the short form together). (B)(i) Someone reading through multiple questions will be learning terms and their acronyms along the way. (B)(ii) Learning in this way while reading through multiple questions could help with understanding older questions with unexplained acronyms. $\endgroup$ – A.M. Aug 10 '13 at 2:39
  • 1
    $\begingroup$ I've a number of times seen SPSS users use GLM to mean General Linear Model (with a Gaussian response but combinations of categorical and continuous explanatory variables), reflecting the unhelpful way the menus describe things in that application (and that, astonishingly, a generalized linear model doesn't seem to be part of the base package and hence many users have no idea of its existence). $\endgroup$ – Peter Ellis Aug 15 '13 at 1:57
16
$\begingroup$

Almost any TLA (three-letter-acronym) will have a large number of meanings. It seems to me that in many cases if there is one statistical meaning there will also be another--if not now, then perhaps in the future. Therefore it is foolhardy to assume that any TLA, no matter how conventional, will be correctly understood even from the context, no matter how obvious it may be to the original proposer or would-be respondents.

We should insist that all acronyms be written out (the first time) in all cases. If a question is worth asking, it's worth spending another ten seconds to ask it clearly and unambiguously.

$\endgroup$
  • 2
    $\begingroup$ +1. I have suggested edits to remove acronyms the first time they are cited, and I think Glen_b's suggestion of including this hint on the help center page would help to support this idea. Is this possible? $\endgroup$ – Andre Silva Aug 7 '13 at 13:34
  • 2
    $\begingroup$ I'd divide this into two. Ambiguous abbreviations within titles are the biggest nuisance. Defining and then using abbreviations within questions is part of what most people do it most of the time and very hard to police any way. $\endgroup$ – Nick Cox Aug 7 '13 at 13:39
  • 2
    $\begingroup$ @Nick Might I suggest that (1) editing the titles (where the meaning is spelled out in the question) and (2) asking for clarification (where the meaning is not spelled out anywhere) may be effective, easy actions to take? $\endgroup$ – whuber Aug 7 '13 at 13:43
  • $\begingroup$ One may argue that titles will become too long. For that, it helps to emphasize that in many situations words which meaning are carried by tags do not need to appear in title. Example: How to do something using R ("using R" can be replaced by R tag). Is that right? $\endgroup$ – Andre Silva Aug 7 '13 at 13:48
  • $\begingroup$ Naturally I agree that these are things to do, and I have done both. I am not clear whether and why you might think I disagree. The action point being suggested by @Glen_b is, I think, strengthening the help rather than changing editing attitudes or practices. $\endgroup$ – Nick Cox Aug 7 '13 at 13:51
  • 1
    $\begingroup$ @Andre Silva Yes and no. "Using R" could be central to the question. If it is, then it belongs in the title. (I assume here that the question is still on-topic for CV.) $\endgroup$ – Nick Cox Aug 7 '13 at 13:53
  • 2
    $\begingroup$ @Andre That's a good question that I cannot answer briefly or unambiguously. People use titles to decide whether to read a question. In some cases the question may be so R-centric, for instance, that perhaps signaling that in the title would be wise, even though ordinarily it would be an unnecessary duplication of information in a tag. Titles can also be targets of searches (see the "advanced search tips"), so in some ways they ought to be thought of as syntactically meaningful sequences of keywords. $\endgroup$ – whuber Aug 7 '13 at 13:54
  • 1
    $\begingroup$ @Andre It is possible to edit the topics help page, but the page relevant to asking good questions--which is where such a requirement ought to appear--is not editable (by your mods, anyway). $\endgroup$ – whuber Aug 7 '13 at 20:57
  • 1
    $\begingroup$ @whuber I agree about TLA's but in phrasing my question I was specifically thinking of a longer abbreviation as an example of being a 'standard' one I wouldn't want to interfere with -- "ANOVA". $\endgroup$ – Glen_b -Reinstate Monica Aug 7 '13 at 22:47
  • 1
    $\begingroup$ @Glen_b I was trying hard to think of standard acronyms; thanks for that example. In some sense, "ANOVA" no longer is an acronym, just as "snafu" is not. (We might call these "paronyms", $\pi\alpha\rho o\sigma$ "former" + $o\nu o\mu\alpha$ "name".) Moreover, the longer the acronym (or abbreviation), the less likely it is to be ambiguous, whence the focus on TLAs ("two/three-letter acronyms"). $\endgroup$ – whuber Aug 8 '13 at 16:41
  • 1
    $\begingroup$ @Andre Thanks. As a start I have added a new section to that post, headed "Style". Everyone is (of course) free to improve it. $\endgroup$ – whuber Aug 8 '13 at 16:56
  • 2
    $\begingroup$ @whuber I agree that ANOVA is now a word in its own right despite its formation as an acronym; the difficulty is not whether or not it's really an acronym now, it's that I don't know how to suggest to a new user, especially one with limited statistical background, how to tell that ANOVA wil be understood, but ANOSIM probably won't. $\endgroup$ – Glen_b -Reinstate Monica Aug 8 '13 at 21:14
  • 3
    $\begingroup$ @Glen_b That's a great point. Here's a proposal that is only partly in jest: if a word is clearly an acronym or you have to write it out in all caps, please expand it on the first use. That lets users of "Anova" off the hook but obliges users of "ANOVA" to learn what it stands for :-). $\endgroup$ – whuber Aug 8 '13 at 21:26
  • 1
    $\begingroup$ @whuber well it certainly works as a jest, because I laughed. As an actual suggestion ... I don't seem to have a better one right now. The new section of that meta post is good. $\endgroup$ – Glen_b -Reinstate Monica Aug 8 '13 at 23:03
  • 1
    $\begingroup$ @whuber Thanks; & I think the wording's perfect. $\endgroup$ – Scortchi - Reinstate Monica Aug 12 '13 at 14:19

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .