11
$\begingroup$

I know there is the very popular SEMNET mailing list, but this is unmoderated and suffers from extended discussions that become quite heated and personal from time to time. Given the popularity of SEM in the social sciences and psychology, I have been wondering why we have so few questions on structural equation modeling on CV ?

The tag appears on page 4 of popular tags, while is all the way down on page 8.

| |
$\endgroup$
  • 4
    $\begingroup$ I don't use SEM and don't know SEMNET but your description that it's "very popular" seems to be the answer to your own question. Extended discussions can be highly informative and even entertaining. More generally, a forum can have downsides for some people in the field, but nevertheless be the most obvious first port of call. More positively, the existence of a dedicated forum can be a real magnet. In one case I know a lot about, a dedicated Statalist seems to attract more Stata questions than all other forums put together. Not everyone likes its policies and styles, but what else is new? $\endgroup$ – Nick Cox Jul 9 '16 at 17:24
  • 1
    $\begingroup$ It feels to me like we get a lot of posts about structural equation models (though I don't know that they all use the tags you looked for). Maybe they don't get enough attention here to attract more questions. $\endgroup$ – Glen_b Jul 10 '16 at 3:09
  • 2
    $\begingroup$ I would say that we have lots of questions about latent variables but they aren't tagged in any consistent way. As about SEM, I guess it is simply not that popular approach -- it is in some areas but not in general (like IRT modeling is something very popular in some branches of psychometrics but not in general and there is not that much IRT questions on CV). $\endgroup$ – Tim Jul 10 '16 at 18:06
  • $\begingroup$ There are software specific resources too, for instance for MPlus and OpenMx, so maybe the questions go there? $\endgroup$ – mdewey Jul 11 '16 at 8:40
  • 2
    $\begingroup$ I don't know if it's the sole reason, but I'd certainly second the point made by @mdewey. Bengt and Linda Muthen, both influential figures in the SEM world, are very active on the message boards for MPlus, one of the most widely-used pieces of SEM software. In fact, I'd suspect that for some users SEM and MPlus are synonymous. $\endgroup$ – Ian_Fin Jul 12 '16 at 13:20
  • $\begingroup$ @Ian_Fin while I agree that the statmodel forum can be good, it can also be very frustrating with many threads ending in "please email us your data and we'll take a look and reply privately" $\endgroup$ – Robert Long Jul 12 '16 at 17:28
  • 1
    $\begingroup$ I take on board all the comments but I just feel that a forum for talking about SEM outside the confines of any software-specific forum is needed. Semnet is supposed to be that but I know several people who have left due to acrimonious exchanges. $\endgroup$ – Robert Long Jul 12 '16 at 17:31
  • $\begingroup$ There are a lot of questions on it indirectly. $\endgroup$ – user46925 Jul 16 '16 at 15:04
10
$\begingroup$

This community is user driven, and our most popular questions are asked by users of all experience levels and applied areas. Meta-questions like yours have been asked: "Why don't we have more questions on X?" And the answer is very obvious: Users of this site aren't coming here with curiosity/knowledge about X. The SE platform synergizes topics: questions that get frequent answers, upvotes, edits, and comments are exponentially more "bumped" than they would be in say forums (perhaps less so than in Reddit).

That doesn't mean the community is unable to support questions tagged or related. SEM is conceptually easy to do, but challenging to implement. The open-source alternatives are shoddy and can be tough to use, lavaan is perhaps one exception; OpenMX is almost impossible. M-plus is non-free software with a vibrant support community. The Muthens have provided detailed examples and discussions, and they frequent their own board answering forum questions fairly politely. Paid software users like to utilize the support services they have exclusive privilege to. On those boards, syntax and conceptual questions are on-topic.

In contrast, a large proportion of CV users are R users. While we now close strict R questions, a majority of questioners use R code or data to illustrate a problem and are looking for R code in an acceptable answer. Most of the time, answerers can fire up an R session and load that code, example data, and provide new syntax or plots to illustrate problems. R simulations to theory questions are frequently accepted answers here. To a lesser extent this is also true of Python. Contrast this with questions that use/illustrate a problem with Stata or SAS code: they go virtually unanswered.

I suspect the community will be challenged by lack of access to M-plus, even in questions not directly about M Plus, and the popularity/accessibility of these questions will be somewhat low.

| |
$\endgroup$

You must log in to answer this question.

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