0
$\begingroup$

Can my question be moved to CrossValidated please? I have flagged the question on SO and asked it to be moved here.

Reason: the package author is active on CV but not SO. He has kindly given some advice about this issue in comments another question by me on CV but I fear if I post a new question on CV it might be closed as a duplicate/cross post.

$\endgroup$
3
$\begingroup$

No offence, but the only valid reason for migration is when one question is off-topic on source SE and on-topic on the target SE.

If you want to reach package author directly, it is a better idea to send an e-mail than to chase him across SE network with off-topic posts.

$\endgroup$
  • $\begingroup$ Thanks, I understand what you are saying, but I don't think it's completely off-topic here as it may be a statistical problem to do with the data itself rather than a programming problem. I have tried emailing him before, but didn't get any response. He seems happy to comment on the problem in the comments on the unrelated question on CV. Is it a crazy idea to migrate it here and hope to get his input and later migrate it back to SO if it turns out to be a programming problem ? ;) $\endgroup$ – Robert Long Feb 2 '13 at 15:36
  • $\begingroup$ Robert, it all depends on what the question is. The one you reference very clearly is off topic here and on topic at SE, because it asks for R techniques to diagnose a crash. If you do have evidence that this is data-related and can introduce additional information about exactly how the data might be causing such a problem, you might be able to put the question in a statistical manner, which would make it on-topic here. $\endgroup$ – whuber Feb 2 '13 at 17:20
  • $\begingroup$ @whuber I do have evidence that it is data related in the sense that it crashes with certain subsets of the data and not with others (I think I mentioned that in the post) and I am working on generating small enough anonymised datasets that can be posted [related question: is there a recommended way to host data which is too big to use dput() ?]. Also, I wasn't specifically asking for R techniques to diagnose the crash - just generally "how can I diagnose the problem". $\endgroup$ – Robert Long Feb 2 '13 at 17:39
  • $\begingroup$ In this case it is difficult to separate the R programming issues with diagnosis from any statistical issues. What statistical techniques of diagnosis might you have in mind? $\endgroup$ – whuber Feb 2 '13 at 17:40
  • $\begingroup$ @whuber possibly issues to do with small cluster sizes and/or specifying the predictor matrix (off the top of my head) but I was hoping Steph van Buuren would be able to weigh in. To be honest it doesn't seem to be a problem with R programming because it runs fine with exactly the same code (and data structure) on other subsets of the data. $\endgroup$ – Robert Long Feb 2 '13 at 17:46
  • $\begingroup$ I see you're getting somewhere interesting, but how are we supposed to help with this if you don't make these concerns explicit in your question and unless you provide sample data? $\endgroup$ – whuber Feb 2 '13 at 18:00
  • $\begingroup$ @whuber I am going to attempt to provide sample data shortly in a major edit to the question. The problem is that it's far to big to use dput() - is it OK to host it externaly ? I fear that if I cut it down so far that I can use dput() (that is assuming it still crashes on such a small dataset) it won't be representative of the actual data. I have now managed to obtain a small-ish sample (~2800 obs on 17 vars) which still crashes and I just need to completely anonymise/jitter it. $\endgroup$ – Robert Long Feb 2 '13 at 18:18
  • $\begingroup$ Yes, hosting externally is a good idea. $\endgroup$ – whuber Feb 2 '13 at 18:28
  • $\begingroup$ @whuber OK - done... hope it's an improvement... $\endgroup$ – Robert Long Feb 2 '13 at 19:37

You must log in to answer this question.

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