My answer to this question was removed on the basis that I didn't answer the question. The question asked to address the following statement:

There is a correlation between the number of people in a country who are Muslim and the number of terrorist attacks

My answer directly addressed this by providing citations which suggest that a correlation does exist.

I also provided a US case study as a further example of correlation, which was further cited as a reason for deleting the answer, although I'm not sure why.

In this instance the moderator action is a subtraction from the quality of discourse. If they hadn't intervened and simply commented, I would have had the opportunity to address any concerns and improve my answer.

Please undelete my answer. Moderators not liking a response is not a valid reason for deletion (that's what voting is for!).

  • While I know it's not your intention to circumvent our system, deleted answers are only visible to a subset of users for a reason. Please don't repost them here.
    – Sklivvz
    Jul 20, 2016 at 9:49
  • LOTS of answers to that question have been deleted, including one which actually calculated the strength of correlation using standard statistical methods (though the methodology was questonable, e.g. it used counts not rates so population was a confound). It'd be nice if the mods could maybe put a comment under that question outlining the requirements for an answer and linking to the relevant policies (e.g. original research) so people don't waste their time Jul 23, 2016 at 10:41

3 Answers 3


To address one of your concerns

If they hadn't intervened and simply commented, I would have had the opportunity to address any concerns and improve my answer.

You still have the opportunity to address the concerns, and improve your answer. The only difference is that while you're doing so, in a deleted state, you will not accumulate negative votes.

  • I was accumulating positive votes until my answer was deleted.
    – quant
    Jul 21, 2016 at 19:11

I've reviewed the deletion. While I disagree with the other moderator that your post does not answer the question, I think it has other big issues that you need to address before I am willing to restore it:

  • None of your sources address correlation explicitly, but you conclude that "she's quite possibly right". You need to support that with a reference.

  • The middle three paragraphs provide some relevant numbers about the number of terrorist attacks and number of Muslims. This is interesting, but to infer correlation from those number would be original research, which we don't allow.

In practice, both the claim and your answer seem to be faulty of cherry picking - any research in the matter must be systematic in nature, because the claim is generalized - and of false precision - without estimating the errors involved, you can't conclude that one number is greater than another with any confidence.

We expect answers to source an expert to deduce a correlation. The purpose of the site is not to perform independent research and review it, but to represent the overall current empirical knowledge on a subject.

On a personal level, I think that you can certainly improve your answer and make it a good one. Best of luck!

  • I have made changes to my answer to make the correlation clearer. I am confused by the definition "correlation" employed on this website. I think the moderators here don't accept evidence of correlation unless its linearity is defined. This is not correlation (it is a subset). Other than the whim of moderators, how am I supposed to know which particular brand of correlation I need to satisfy in order for my answer not to be removed. In my particular case I showed correlation, but did not bother to identify its linearity (because the question didn't ask for it).
    – quant
    Jul 23, 2016 at 5:05
  • @quant there is only one way to show correlation on this site: find evidence that addresses the question directly and shows correlation. This does not include inferring correlation from evidence. Think about a question like "is there a correlation between aspirin and ulcer?": you'd answer that by finding studies on the specific claim, not by trying to show the correlation between aspirin usage statistics and ulcer statistics.
    – Sklivvz
    Jul 23, 2016 at 10:04
  • I'm not convinced that you (and apparently other mods) understand what correlation is, or at least you have a very specific definition of it that is not related to common usage. Your example is a case in point; a relationship between aspirin usage and ulcer statistics would demonstrate correlation (by definition), but not causation. You're clearly after causation. The question did not ask about causal relationships. I don't know how I can be clearer about this, and it frustrates me that I am forbidden from answering a question simply because the moderators do not understand it.
    – quant
    Jul 23, 2016 at 11:18
  • @quant I am only genuinely trying to help you. I understand correlation. I understand that it is not causation. I understand that correlation can be an arbitrary function. Still it is our requirement that you let an expert infer whether the correlation is present and not attempt to determine that yourself.
    – Sklivvz
    Jul 23, 2016 at 11:27
  • sorry about my tone - I do appreciate you trying to help. I disagree that it is in the spirit of the rules of this site though that expert sources need to be cited to show correlation in instances where correlation is trivially observable from the evidence (in this case to the point of being tautological). I think I've made my point so I'm happy to let this lie. I hope you can see my frustration though - I won't be taking the time to post answers here again.
    – quant
    Jul 23, 2016 at 12:13
  • 1
    @quant OK, sorry to hear you are frustrated! In order to show a correlation, were we to allow it, it would certainly not be enough to look at the top spots on the terrorism index. A correlation needs two data sets, and needs to be shown to exist for the whole set and not some choice spots. It is trivial to make mistakes in these matters which is precisely why attempting research answers has never worked well on this site.
    – Sklivvz
    Jul 23, 2016 at 12:57
  • We're now going in circles so I don't think there's much to be had by continuing this discussion. I think I have a better understanding of what is expected in terms of answers on this site, so I appreciate the time you've taken to respond.
    – quant
    Jul 23, 2016 at 13:05

I'm disappointed that your post here suggests that the post was deleted purely because I "didn't like it", rather than directly addressing the explicit issues with the answer that were highlighted in a comment.

My answer directly addressed this by providing citations which suggest that a correlation does exist.

Your answer did not directly address this. All it did was provide citations which suggested the correlation might exist. In fact, the closest it came to a conclusion was to say "So she might be correct [...]". It was hardly a definitive response. You did not answer the question.

The strongest argument you gave was a "quick glance" (your words) at one end of the scale of terrorism and noting many of them were populous with Muslims. This is cherry picking. If the lower end of the scale also includes countries with a large number of Muslims, the correlation may be zero or negative. That section of your answer should be removed.

I also provided a US case study as a further example of correlation, which was further cited as a reason for deleting the answer, although I'm not sure why.

Using the US-only data - that US Muslims extremists were disproportionately more likely to commit a terrorist act than the general US population - is highly problematic. Not only does it not answer the question (which was about correlations between countries, not inside the USA), it suffers from the fallacy of hasty generalisation. If terrorism is committed by disenfranchised minority groups, suffering from social and political injustices, and if Muslims tend to be a disenfranchised minority group in the USA, latching onto the religion of (a disproportionate number of) terrorists and assuming that applies globally is inappropriate and inflammatory. That section of your answer should be removed, or replaced with a citation of a rigorous analysis.

Once these sections are removed, there is nothing left in the post that answers the question.

I would be happy to see this post fixed and reinstated. The requirement for a good answer here is straightforward, but at the same time I think it will be difficult to answer. We need a simple statistical analysis showing whether the correlation coefficient between a count of terrorist acts (per capita?) per country, and a count of Muslims (per capita?) per country is positive. The hard part is finding an appropriate expert who has bothered to do the calculations so they can be quoted here.

  • I have edited my question to hopefully address some of your concerns. I think you have set an unrealistic (and, frankly, incorrect) expectation for the burden of demonstrating correlation. Correlation requires a relationship between two variables that fails to exhibit probabilistic independence. If countries with higher incidences of terrorism also exhibit higher fractions of muslim populations, correlation is demonstrated. Whether or not it is linear, convenient, relevant or causal is simply not relevant to the question.
    – quant
    Jul 23, 2016 at 4:48
  • Regarding the cherry-picking comment. It would be cherry-picking if I used this information to make claims about causal relationships (e.g. Islam is violent, Muslims are terrorists, etc.). I very explicitely did not do this. What happens on the lower end of the GTI scale is simply not relevant. If a trend were visible at the lower end, it would simply serve to qualify the nonlinearity of an already proven correlation.
    – quant
    Jul 23, 2016 at 4:49
  • Regarding the US data; I think you are right, and have removed it.
    – quant
    Jul 23, 2016 at 4:50

You must log in to answer this question.

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