This question was just migrated to stackoverflow (well, an hour ago).

As far as I can see a great deal of it's pretty solidly on topic here.

"What is this function called?" is not a SO question, it's a CV question.

"Is this function linearizable?" --- there's no way that's an SO question. It's a stats question. How are SO people to be expected to talk about the fact that it depends on how the error enters the model?

"How do I estimate the parameters?" (paraphrasing) is a statistics question.

Those are right in the title.

There's code in the question, but the code related parts are not central to the statistical aspects, and the existence of code -- or even a question about code -- doesn't mean the statistical aspects don't belong here. At most I think the OP could have been encouraged to separate some aspects of the question off, but we have no business sending stats questions to SO.

Anyone involved, please take a look at my answer, and explain to me how a question which has an answer like that belongs anywhere but here -- because if that's our standard, I clearly have no clue whatever how any of this works.

Please help me understand the basis for this choice. I really don't understand how those questions at least don't belong here.

(I happen to feel that some of the remainder also belongs here, because it also contains some statistical aspects, but the parts I mentioned should be sufficient.)

  • $\begingroup$ Can one see, after the migration happened, who voted to migrate it? With closed questions this information is always displayed. $\endgroup$
    – amoeba
    Oct 16, 2015 at 9:49
  • 2
    $\begingroup$ @amoeba I knew exactly who migrated when I posted but I don't want to point the names out; because then it's too much like pointing fingers. If people choose to respond themselves, that's great but I am not going to raise any names myself; I think that would be counterproductive. $\endgroup$
    – Glen_b
    Oct 16, 2015 at 10:30
  • $\begingroup$ Of course, I understand. My question was if non-moderators can see this information somewhere; not in this particular case but in general. It looks like we cannot and it's a bit strange (because for closed questions or the ones put on hold, this information is right there). $\endgroup$
    – amoeba
    Oct 16, 2015 at 10:35
  • 1
    $\begingroup$ To my knowledge, I have no easy way to find out what you see if I don't already know. I have asked that mods be able to set their privilege level to be lower temporarily but it's a dupe and 'status declined'.. $\endgroup$
    – Glen_b
    Oct 16, 2015 at 10:41
  • 6
    $\begingroup$ +1 On a positive note, this event is a milestone of sorts. Until a couple years ago, it was rare to find five non-mods with enough reputation to close or migrate a question. I am pleased to see a wider cross section of the community involved in the review and maintenance needed to keep our site in good shape. $\endgroup$
    – whuber Mod
    Oct 16, 2015 at 12:52
  • $\begingroup$ @amoeba, anyone w/ rep >10k can find out who voted to close. You go to the review queue then click on tools -> migrated -> away, find the thread in question & click on it, which will take you to the stored, original CV version w/ the closed by ... < reason > information. $\endgroup$ Oct 16, 2015 at 14:49
  • 3
    $\begingroup$ Thanks, @gung! As I understand now, one can also append ?noredirect=1 to the end of the URL address; then the redirect to SO won't happen and one can read the "migrated by" information. This is actually available to all users, independent of the rep (I've just checked after logging out), given that they have the original URL. $\endgroup$
    – amoeba
    Oct 16, 2015 at 16:46
  • 1
    $\begingroup$ @amoeba There's actually no need to do manual URL changing; the message on the stackoverflow page, "migrated from stats.stackexchange.com 12 hours ago", in fact has a link to the CV version of that post even though the way it's written makes it look like it's just a link to the CV homepage. $\endgroup$
    – Danica
    Oct 16, 2015 at 20:08

1 Answer 1


Note: I was one of the close voters.

You are right that those questions are in the title. However, I interpreted 'how to fit this model in R?' as the real question. (According to my reading, the central sentence is "I can fit the model using nls without a random effect, but am having trouble getting the model to converge".) Things like 'what the name of the function is' could certainly stay here, but I interpreted most of that, including "can it be linearized", as looking for leads that would help the OP fit the model. The OP listed some R code, and (while acknowledging the possibility of other software) would clearly have preferred a solution to fit the model in R.

The suggestion was made by someone who I suspect is very knowledgeable about these issues that the OP would be able to get the help he needs on Stack Overflow (which is my priority). Therefore, I voted to migrate.

I continue to hold that if the OP gets that help and is able to fit the model, the main issue will have been addressed, even if he never learns the function's name or if it could be linearized.

I respect the fact that people can have a difference of opinion on this. I have always believed that the criterion for what site a given thread belongs on depends on what the OP needs explained. This will always be an interpretive act on the part of the voter, sometimes more than others. For example, there have been several occasions where I have voted not to migrate a question about R code, because I could see that the reason the OP had used the code incorrectly was because they misunderstood the underlying statistical issues. Thus, explaining the code or correcting it would not have been a real solution; the OP just needed to understand the ideas and they would have been able to generate correct R code on their own. Trying to figure out what an asker really needs / wants to know is tricky and people will occasionally get it wrong.

  • 6
    $\begingroup$ It's not entirely obvious though, at least to me, that failure of a non-linear mixed model to converge is just a programming issue. There's nothing wrong with the OP's nlme code. (I admit I'd prefer a good answer about convergence problems on CV to the OP's being given lme4 code that works for some reason, even if that made the OP happy.) $\endgroup$ Oct 16, 2015 at 15:14
  • 5
    $\begingroup$ Thanks. I suppose it's two points: (1) that statistical computing's on-topic, even if the minutiae of how to do $X$ with software $Y$ aren't, & (2) that explaining principles trumps immediate gratification of the OP; both in my opinion. $\endgroup$ Oct 16, 2015 at 15:50
  • 6
    $\begingroup$ +1 for stepping in and giving an explanation about your point of view. $\endgroup$ Oct 16, 2015 at 16:00
  • $\begingroup$ Those are all reasonable points, @Scortchi, but I would still vote to migrate here. I don't think SO has to just have try <some code> answers; eg, understanding coding strategies & problem solving should be part of their mandate too. I don't really see the Q as about statistical computing any more than it's about regression, even though it has a regression model. Your 2 is tricky: I think we want both, & even if we emphasize principles over helping, it needs to be the principles that are really behind the Q. My interpretation remains that this one is ambiguous, but that I lean towards SO. $\endgroup$ Oct 16, 2015 at 17:18
  • 7
    $\begingroup$ +1 Thanks for the response gung, I appreciate you giving the reasoning. I disagree with much of your assessment (so the +1 isn't to signify agreement in this case, but an appreciation of a thoughtful answer). I'd like to try to find a way to achieve better consistency on this sort of question. If I saw that question there, I'd vote to migrate it here (and many people on SO migrate much less clearly statistical questions than that to us). $\endgroup$
    – Glen_b
    Oct 17, 2015 at 1:58
  • $\begingroup$ -1. I downvote because, after reviewing this admittedly old discussion, I think the Q in question should not have been migrated. It's a pity there was no opposing answer at the time; @Glen_b's comment above saying that the question belongs here got 7 upvotes whereas your answer got 6 (before my downvote) and at least 2 of these upvotes are not because of agreement (see comments above). So clearly no consensus has been reached here, but it seems that the majority of Meta voters thought the Q should have stayed. $\endgroup$
    – amoeba
    Sep 17, 2016 at 22:52

You must log in to answer this question.

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