We got a few questions about probability distributions, e.g. here and here.
Are those questions on-topic for the OR stackexchange?
$\begingroup$
$\endgroup$
1
-
2$\begingroup$ Some of these may be better suited for Cross Validated stats.stackexchange.com , but I wouldn't close them as off topic at OR SE. What about questions on optimization applied to science and engineering rather than OR? If the focus is on the science and engineering, perhaps better elsewhere, but if on the math or computation of optimization, even if not "O.R", then why no take advantage of the large pool of optimization experts at OR SE? And I wouldn't downvote any of these questions, unless thee are of the "someone do my homework for me" or similar type. $\endgroup$– Mark L. StoneCommented Jun 17, 2019 at 14:13
Add a comment
|
3 Answers
$\begingroup$
$\endgroup$
8
I think questions purely about probability distributions with no reasonably obvious link to OR should be closed as off-topic and/or moved to the Stats or Math Stackexchange.
Questions with a link to OR (e.g. inventory theory, part of a model, simulation, modeling, ...) can be on-topic, if they require OR expertise to answer. A pure-stats question which just happened to appear while working on a OR related problem should be moved to the stats/math stackexchange.
Of course there are questions that could meet the requirements for both stackexchanges. In such a case I would expect the OP to search the stats/math exchange before asking on the OR exchange.
answered Jun 14, 2019 at 8:52
-
4$\begingroup$ I would use this rule for questions about any mathematical object at all. If there is no connection with OR, it is off-topic, and otherwise it is ok. Of course, the connection may not always be clear, but in that case it is up to the owner of the question to clarify the connection. $\endgroup$ Commented Jun 14, 2019 at 9:16
-
4$\begingroup$ And that OR connection should be materially relevant to the question in a way that needs specific OR expertise to answer, and not just a coincidental connection where the author just happened to be working on an OR project when that [math | stats | academic] problem came up. The author may have encountered these problems while engaging in OR-related activities, but such questions are not directly related to the subject of operations research itself. (I call this the "favorite soft drink of programmers" problem) $\endgroup$ Commented Jun 14, 2019 at 13:20
-
2$\begingroup$ I have some concerns over the wording of this answer but welcome the discussion. I've posted some thoughts and welcome further discussion. In the meantime, I will add some context to the Q/A to clarify the [OR] connection. $\endgroup$ Commented Jun 14, 2019 at 14:11
-
1$\begingroup$ I changed the answer a little bit to reflect the comments. I agree that OR expertise should be required to answer the question. I am not quite sure how moving questions to a different stackexchange works, but I think it is reasonable to move any stats questions that does not need OR knowledge to answer to the stats exchange. $\endgroup$ Commented Jun 14, 2019 at 15:53
-
1$\begingroup$ I appreciate your discussion on this. This seems tricky in practice due to (at least for now) the uncertainty around what the community considers "OR Knowledge." As you have reworded your answer, we seem to agree that OR is more than Optimization, but what exactly is "OR knowledge"? $\endgroup$ Commented Jun 14, 2019 at 16:33
-
$\begingroup$ @SecretAgentMan What exactly is "OR knowledge" is is something none of us will find out and never agree on. But you are in good company, because this most certainly is the case for Computer Science as well. At the very least, the question whether a mathematical construct is within the scope of algorithmics is non-trivial. Take the notion of treewidth, for example. The definition itself appears to be completely unrelated to algorithms, yet due to the fact that its main application is within algorithms, I think most people would agree that it is on-topic for CS. $\endgroup$ Commented Jun 16, 2019 at 11:48
-
$\begingroup$ @SecretAgentMan So, what happens in practice on Computer Science is that we find out whether something is "CS knowledge" or not by having a few broad strokes (e.g. algorithms are on topic, implementations generally not. Programming language design is on topic, programming language usage is not), and figuring the rest out on a case-by-case basis, by closing/reopening and discussing the situation when people disagree. I can say that this works reasonably well on Computer Science, and I don't see any particular reason why that wouldn't work here. $\endgroup$ Commented Jun 16, 2019 at 11:52
-
$\begingroup$ Thanks @Discretelizard, I'm all for the process. $\endgroup$ Commented Jun 17, 2019 at 2:41
$\begingroup$
$\endgroup$
2
I mostly agree with Michael, but I think this will be a tricky line. E.g., I would like to see the one about loss functions considered on topic; it seems sufficiently related to inventory theory.
Also to Discrete lizard's comment, I think in enforcing whatever general policy we land on for math-related questions, we need to remember that folks have different definitions of what OR is and be a bit generous.
-
2$\begingroup$ Yes, maybe we should have a more general meta question: what to do with maths questions that have an OR connection but do not require OR knowledge to answer? $\endgroup$ Commented Jun 14, 2019 at 16:01
-
3$\begingroup$ +1 for “being a bit generous”. I think the fact that OR.SE will be substantially smaller than Math.SE (due to the inherent nature if the subject matter) affords us more flexibility when enforcing “on-topic” rules. $\endgroup$– David M.Commented Jun 16, 2019 at 22:09
$\begingroup$
$\endgroup$
5
This is a good discussion to have -- I've upvoted the question despite one of the linked questions being mine. I agree with @E.Tucker that this will be a tricky line given the breadth of OR.
I almost didn't post that Q/A but decided it decidedly fit for the following reasons (of course, the community may disagree). OR is about supporting decisions --- using data, models, etc, but at the end of the day it supports decision-making. It may be process improvement, but still decisions.
On the site so far, I've seen inventory and operations management questions. I've seen (& even hesitantly participated in) reference questions and "softer questions."
There are Optimization questions which clearly fit. However, Simulation is another pillar method of OR. I wrote my Q/A in the spirit that it was a modeling decision when conducting input modeling for a Simulation or another Stochastics Process model, which I feel is within the OR scope.
I'm a bit concerned over this answer's wording (edited after this answer posted) that questions with no "obvious link to Optimization and OR should be closed" may unintentionally imply to others that OR is about Optimization. That's partly true. But OR is much broader than that. For example, in the US Military, Simulation is used more than Optimization by OR practitioners based on multiple surveys.
Again, this is a good discussion to have. We need to think carefully about what is on-topic at a site for OR.
-
4$\begingroup$ IMO simulation is solidly within the scope of OR, and OR.SE. I can't speak for others, but if there were objections to your Q&A about the beta distribution, I suspect it was because the connection to simulation was not clear— not because questions with a clear connection to simulation would be off-topic. $\endgroup$ Commented Jun 14, 2019 at 15:03
-
$\begingroup$ I have updated the Q&A to include more context. $\endgroup$ Commented Jun 14, 2019 at 15:09
-
2$\begingroup$ Yep, it works for me. I feel that question is on-topic now. It's a little on the border -- a question that could be asked here or could be asked elsewhere, with emphasis on the "could be asked here part." $\endgroup$ Commented Jun 14, 2019 at 15:11
-
$\begingroup$ While I could totally live with this answer, I reworded my answer to be stronger to stimulate discussion. I can live with stats questions (given some OR reference), but I can also support moving questions that do not require OR knowledge to the stats exchange. $\endgroup$ Commented Jun 14, 2019 at 15:58
-
$\begingroup$ I appreciate your discussion on this. This seems tricky in practice due to (at least for now) the uncertainty around what the community considers "OR Knowledge." As you have reworded your answer, we seem to agree that OR is more than Optimization, but what exactly is "OR knowledge"? $\endgroup$ Commented Jun 14, 2019 at 16:33