The question "Gaussian kernel density estimation with fixed covariance (with python)" was incorrectly migrated to StackOverflow.

Can we have it restored to Cross-Validated?

Based on this meta question it does not appear as though the original post exhibited the negative criteria listed in the highest-voted answer:

  • The question includes tags for things besides a programming language.
  • The question could be easily re-stated in English (with no computer code) since the idea that there is just a simple programming option to enable a full covariance-like bandwidth parameter to KDE is not a coherent criticism (there is no such simple programming option and it would involve novel research to create one.)
  • The question is not purely mathematical, since it is about data-driven or otherwise covariance based choices in modeling data with KDE process and presumably applies to actual data given the programming references.

Along these lines, from the meta-question I linked above we have:

"Something to notice about most of these examples is that their titles contain technical statistical terms. The presence of such words does not imply the question itself is on topic!"

I am curious if there is community support for the analagous claim: "The presence of programming-language-specific words, tags, or phrases in the title of a question does not imply the question itself is off topic!"

The question does involve a programming language, but as I interpret it, the question asks about a method to employ a more "covariance-like" notion of bandwidth within the Kernel Density Estimation framework.


I voted to migrate this question after some thought and reviewing the comments; this was not a clear or automatic case. I ultimately relied on the original text to make the decision for me: it explicitly asks for a Python solution and the OP takes pains to intimate that he could program the algorithm himself, but wishes to avoid it. That was a sufficiently clear indication that this is a pure programming question and leaves me comfortable with the decision. At this point it's out of our hands: you would need to appeal to the SO mods either to reject the migration or migrate the question back here. You ought to wait for clarification from the OP first.

  • 1
    $\begingroup$ This is a good point. I had not considered the point where the OP mentioned being able to program it but wished to avoid it. $\endgroup$ – ely Sep 11 '13 at 13:57

I also voted to migrate that question. My criterion is simple: What does the OP need explained? If it's a statistical concept, then the thread belongs here; if it's something about how to use software, then it belongs elsewhere. The OP asked how to fix the covariance of a Gaussian kernel to some pre-specified value in Python. A perfectly satisfactory answer could have been:

Yes, that's easy to do. You set argument=value.

  • 1
    $\begingroup$ (+1) I also voted to migrate the question using similar logic. $\endgroup$ – Macro Sep 10 '13 at 23:49
  • 2
    $\begingroup$ I appreciate this perspective, but I'm of the opinion that an answer along the lines of "Yes, that's easy to do. You set argument=value." would specifically not be a useful or adequate answer to the question. But I agree the OP seems confused about that (based on whuber's observation above) and so based on the OP's intention, I suppose migrating it is fine. $\endgroup$ – ely Sep 11 '13 at 14:00

You must log in to answer this question.

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