I asked a question about how to obtain closed-form solutions for a particular set of three equations.

invert multinomial logit link with three unknown

The question was placed on hold because it was not clear to some what I was asking. So, I edited the question to clarify.

In my question I provided a way to obtain closed-form solutions for a set of two equations and explained I could not do so for a set of three equations. I posted R code showing how far I was able to get (and added some R code as I made incremental advances after posting the question).

Eventually I was able to derive the closed-form solutions for the set of three equations in my question. In an answer I provided those solutions as well as the R code I used to obtain them.

Nevertheless, my question was closed, and down-voted and my answer was down-voted. The answer was down-voted even though my derived closed-form solutions appear to be more accurate than estimated solutions provided by multinomial logistic regression. (Although, I did not provide such a comparison in my answer.)

Now someone has voted to delete the question.

This is all very confusing to me. I do not know how to make the question clearer. The question and answer appear, to me, to be entirely mathematical and perfectly suitable for a mathematics forum, and have abundant utility for further exploration of multinomial statistics (by allowing a comparison of 'truth' vs estimation).

Why is such a question and answer considered unsuitable for this forum?


I now see that my correct solution has been deleted.

  • $\begingroup$ There is a dedicated thread for asking for reopen/undelete votes; Please post at math.meta.stackexchange.com/q/19042/9003 $\endgroup$
    – amWhy
    Jan 7, 2018 at 15:45
  • $\begingroup$ "I now see that my correct solution has been deleted." Are you talking about your answer? It's still there and was never deleted. $\endgroup$
    – JRN
    Jan 8, 2018 at 1:00
  • $\begingroup$ I can see my answer but it is faded out the way that deleted but still visible posts are. I will get over it. I came here because I needed the solution. I have the solution now. I thought it might help other people, but this is just too much drama for me. Enough to last several years. $\endgroup$ Jan 8, 2018 at 1:03
  • 6
    $\begingroup$ Your answer is faded out because it has a net of three downvotes. That has nothing to do with deletion. $\endgroup$ Jan 8, 2018 at 2:35

1 Answer 1


Here are two big problems with your post.

Wall of Code

You purport to post the closed form solution to a problem, but that's not what your post contains. Instead, you post a wall of source code that is mostly irrelevant (I think most of the script is about empirically verifying your algebra work — you never actually explain!) and somewhat obfuscated (the actual equations are buried in the programming logic one needs to correctly compare floating point values).

The readability of your post would increase 700% if you just posted the specific formula you say you are posting. And it would be even better if you used the mathjax capability of this site to format it as mathematics rather than as a source file.

Where's the question?

Your actual question is buried in-between the two walls of code. I bet most people who looked over your post never even noticed it.

It's your responsibility as the author to make your question easy to find. Ideally the reader should understand your question shortly after starting to read your post, or at least be put in the right frame of mind to know where you're going with your exposition.

(the question might be fine where it is if you correct the problem mentioned in the previous section)

  • $\begingroup$ The answer does contain the closed form solution to a problem. I will move the question up, use the formatting ability of the site and remove the code or most of the code. $\endgroup$ Jan 7, 2018 at 7:47
  • 1
    $\begingroup$ @MarkMiller "The answer does contain the closed form solution to a problem" Actually it does not. Already the first identity in it is wrong. which posits that $\ln((1-a)e^x)=(1-a)x$. $\endgroup$
    – Did
    Jan 7, 2018 at 17:22
  • $\begingroup$ @Did If an equation is wrong, the mistake was introduced in revising the formatting. The original formatting was code that actually ran and was correct. The original formatting is still visible under the edit history and I can check it against the new formatting. $\endgroup$ Jan 7, 2018 at 20:54
  • $\begingroup$ @Did I believe I have corrected the typos in the first two log equations of the question and the first three log equations of the answer. These typos arose in my confusion from trying to learn the formatting used on the forum. $\endgroup$ Jan 7, 2018 at 21:35

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