How to ask a "good" question on CrossValidated?

Question

Background

First of all please note that I have been a user on SE sites for over two years now; more than 20 months on CV. While active on StackOverflow and Sports.SE I never quite managed to get going here, despite my best efforts. It is clear to me that I am experiencing a communication problem with members of the community, perhaps due to the fact that I am not a statistician by education.

Please take a look at my previous questions (there are only 8), and you can perhaps see that I have been trying best to explain my background and question as clearly as possible. That being the case, more often then not, I get questions in return, and sometimes even dismissive remarks. In most cases questions end up turning into endless swamps; more I try to salvage something, more I get stuck on comments/questions that lead nowhere. To make matters worse, I seem to have similar problems with (bio)statisticians I meet at conferences. Needless to say, I find this very frustrating.

Question(s)

Based on the background I have mentioned above, I realized that I am here to learn and improve on my half-forgotten understanding of statistics. So I would really appreciate if you could help me phrase my questions better, so that I can hopefully get more useful answers in the future, or better yet, so that I can make better use of the answers I get to my questions. (Note that this is an attempt to make a good FAQ entry for new members as well, we are currently trying out something like this for Sports.SE.)

• How can I make it clear that my understanding of statistics is not complete (i.e. I might be experienced in some subjects but might have completely forgotten/misunderstood some fundamentals)? This is important to avoid being overwhelmed/confused by the answers from more advanced members of the community.

• How can I make sure that the answers/comments I get are to the point of what's being asked? (more often than not I do not see the relevance of the comments/questions I get to my questions) Misunderstandings occur often, and are usually a major part of the problem, IMHO.

• What are some characteristics of a good CV question?

• If the problems that I have been experiencing are due to the fact that I am not a statistician, is this site ONLY aimed at statisticians, or for people that want to get answers to their questions on statistics?

I apologize for the long post and hope that you can provide some insights on this matter.

---

Disclaimer: I wrote and rewrote my title several times, not being happy with the way it sounds... If you don't think that the title accurately reflects the content please feel free to comment/edit.

---

Edit: I got some nice comments, thanks to @whuber and @Procrastinator. I would really like it if the input was given in answers rather than comments, so that I could refer to/comment back to individual suggestions instead of having a long chain of comments to the question at hand...

Answer 1

Some good comments, which partially cover your questions. I'll try to cover them methodically here.

What makes a good Cross Validated question?

In this order, or similar:

• A clear introductory statement of what you are trying to achieve: what point you are trying to get across?
• Why are you doing it? Homework? Part of your job? Self-education?
• A brief, clear description of the data you have. Describe each variable. Ideally, provide a couple of rows of data, so people can see the format it's in. Say what programming language you're using.
• If, and only if, you can see a good way to generalise the question (other projects you plan on doing, using similar concepts or something), then put it out there. Otherwise, just assume that the people who come after you are not stupid and they will be able to generalise it themselves. Everyone is used to doing this, as this is the most common way we learn: by example.
• Clarification of your experience (see next section), but only if you feel it's necessary.
• A clear statement of your question or questions. You are best off sticking to one clear statement of your core question, rather than multiple questions or multiple formulations of the same question. Those questions may seem related to you but there's a chance they won't seem so to others. You can always post follow-up questions to clarify certain aspects. Doing so and linking back to the original question will make it easier for people to pick up what you're trying to communicate. If you think that you have side-questions that aren't suitable as separate posts, then feel free to add them, but make sure to indicate that they are less important. One way you could do this is to make your main question statement bold.
• Re-read your question and remove all guff. You don't need to apologise for your perceived inadequacies nor do you need to thank people for anything. Anything that doesn't add relevant information is only detracting from your communication effectiveness.

Style

There are few stylistic restrictions: write naturally and use a language familiar to you rather than one you think statisticians might speak. Work to make your question stand by itself and be understood by everyone. In particular:

• Define or explain any technical terms and mathematical symbols you use. Do not assume everyone uses your terms in the same way (statistics applies to a huge variety of disciplines) nor that a given symbol invariably stands for one thing.

• Explicitly write out, in full, the first occurrence of any abbreviation or acronym.

• Strive for standard appropriate English usage, grammar, and spelling. It's easy: pay attention to the automatic spell-checker markup when you're writing. You need not be as formal as, say, an academic journal, but don't be too chatty either. Avoid abbreviations or calculated misspellings used in texting (e.g. u or U for you).

• Use the highlighting tools (found immediately above the textbox) to create headers, emphasize text, and inset code and quotations.

• Format mathematical expressions using $\TeX$ enclosed between dollar signs $\$$ (or double dollar signs $\$\$$ for centered expressions).

Some things to avoid:

• Do not beg for help. We know you would like a great answer as soon as possible.

• Do not include identifying information such as your name or e-mail address--we will remove it if it appears, anyway. You can place those into your user profile for all to see.

• Do not solicit offline or private responses: questions and answers are posted here for everyone. We are more than (just) a Q&A site!

Clarifying your experience

Take your lesson here from the master, Donald Rumsfeld:

• Known knowns: State any relevant information that you already understand, as concisely as possible. Say why it's relevant. Feel free to say something about how experienced you are in stats.
• Known unknowns: If you're aware of certain methodologies, theorems, tools, etc. that you think you might need, but are not familiar or comfortable with, say so, and say why you think they're relevant. People can pull you up here, if necessary.
• Unknown unknowns: You don't know about these, so don't fret about them. If someone brings some up, then you can add them to the previous category.

You can't do much better than that. Don't labour over it, and keep it as short as possible, or people will get bored and leave.

If someone mentions something for which you require clarification, ask in the comments. If it changes the question substantially and answers have already appeared, you may be better off keeping the question unaltered (out of respect for the material already posted) and starting again with your new information. Otherwise, you can just edit the question.

Avoiding Misunderstandings

Misunderstandings, in my experience, occur because of lack of context. Statistical concepts are hard to communicate clearly and it's easy for someone reading your question to completely miss what you're trying to get at. A good way to add context is to provide a solid example.

If you don't understand the relevance of a comment, say so, and state clearly why you think it is not relevant, so that the answerer/commenter can get a better handle on where you're coming from. This will allow them to give you more targeted context for their comment.

Do I need to be a statistician?

No. Certainly not. The site is aimed at people interested in statistics, practice or theory. You just need to be open to learning.

It also helps to be a good communicator and to know your audience. This is difficult to learn in any situation, let alone one in which most conversations are unintelligible due to lack of experience (I've been hanging around here for almost a year, and I still don't know WTF most people are talking about;)). Remember that people here come from all different backgrounds (different languages, different areas of study, different levels of experience with stats) and you can't assume that anyone knows what you're talking about.

The great thing about StackExchange is that it's really hard to be a troll here, so no-one bothers much. If it looks like someone's being narky, it's probably best to assume that they're just not communicating well, and ask for clarification. Above all, please be patient and give everyone the benefit of the doubt: we're here to help.

Answer 2

Writing good titles

Mark Harrison's CW answer on Meta Stack Exchange brings a list of useful hints for writing meaningful Stack Exchange titles. These are an expanded version of that guidance.

1. Make the topic stand out.

This is the most important filter on a question. People scan content in the front page very fast. Users need to know what the question is about just by looking at the title. Make sure the necessary keywords are included in it.

For example, let's take a look at the following title:

Making sense of principal component analysis, eigenvectors & eigenvalues

The OP ("original proposer") wants to understand better what is PCA, eigenvectors and eigenvalues. Very simple. In the body he provides more details and arguments about the way he wishes to understand such subjects.

2. Keep it short.

It's not necessary to explain the question in the title, but what it is about without being too broad.

Examples:

Difference between logit and probit models

Is $R^2$ useful or dangerous?

3. Lead with the most important words.

Avoid general openings. For example:

I want to show $E(B(t)-B(s))^4=3(t-s)^2$ could be just:

Show that $E(B(t)-B(s))^4=3(t-s)^2$ (tagged with proof)

4. Make those words useful for searching.

Formulas and generic words add little information. Compare the previous example to:

How to compute moments of Brownian motion?

5. Be specific.

For example, instead of:

Business Forecasting

write:

How to incorporate seasonal patterns in forecasts?

6. Don't start with "How do I..."

Writing in that style ensures that your title will fail criteria 1-3, and get less attention than it deserves.
This is a question site, and people will understand your titles are questions.

7. Don't sweat replicating a tag keyword.

Don't use loose tags in titles, unless they are substantial for its comprehension.

Example of a title where two tags are substantial:

Is there any reason to prefer the AIC or BIC over the other?

In this case aic and bic are essential for the title to be understood. Without them, we'd have:

Is there any reason to prefer AIC?

Although this title is useful, it begs the question "prefer it to what?". Both AIC and BIC appear to be important aspects of the question.

Example of a title where the tag is not substantial:

Regression - volatility vs return

The word "regression" seems to be loose (better use it only as a tag). A better title could be

Estimating the effect of volatility on losses

(tagged with regression, times-series, econometrics, hypothesis-testing, and volatility-forecasting).

8. Don't beg.

The surest way not to get a good or fast answer is to plead that you are desperate or have limited time. Focus on asking a clear, interesting question: the answers will come flying in.

9. Use good English.

Misspellings, poor grammar, typographical errors, and inconsistencies all suggest you couldn't be bothered to make your question presentable. Why should anyone bother to read or answer it, then? We are considerate of people for whom English is not a first language, so don't let your ignorance of grammar or spelling stop you--but at least do your best to check everything as well as you can and try for consistency (an English-speaking friend could be helpful). At a minimum, read your entire question (especially the title) after you have posted it and edit any errors immediately.

10. Avoid abbreviations and jargon where possible.

To keep titles brief, you might want to abbreviate (such as "PDF" for "probability density function"). But please spell out all abbreviations early on in the text of your question. If the title refers to an arcane term or concept, consider defining, describing or explaining it in the leading lines of the question. At least provide a link to a definition.

Want a good answer fast? Make the title interesting to everybody who might be qualified to answer your question. That alone suggests avoiding jargon and arcane terms.

---

Mark Harrison also says:

Finally, be flexible.

Different questions benefit from different styles of titles. Applying any single hard and fast rule is probably a mistake.

---

_{Suggestions about writing good titles specifics to statistics are also welcome.}

Answer 3

Use tags

Tags allow us to sort and filter questions into categories that we are interested in. If you tag your question well, you will increase your chance that an expert in your particular problem will see your question.

When you are done assigning good tags, don't post your question yet. Instead, search through previous questions carrying your tags. You can do this simply by putting all your tags surrounded by "[]" in the search box. Here is an example. Chances are you will find something that already answers your question. (If you do, don't forget to upvote the question and answer so the process improves further.)