When browsing unanswered questions, I very frequently find myself wanting to request the user who's asking a question to provide more context. Specifically, I often want to ask some form of "What are you trying to do?", such as "What is the goal of the analysis?", or basic questions about what the data at hand is like, such as "Where does the data come from?" or "What's the sample size?" or "What's the base rate of the DV?". Here are the first five clear examples I saw by just looking at the current page 3 of the new queue:

  1. For this question: What is "validity" supposed to mean? What are the nodes and edges of the graph supposed to represent?
  2. For this question: What are $b_2$, $b_3$, and $y$? What is the goal of the analysis?
  3. For this question: Where does the data come from? What is the goal of the analysis?
  4. For this question: Where does the data come from? Gung's questions in his comment are also on point.
  5. For this question: Where does the data come from and what is the goal of the analysis? The use of the term "bookings" gives the reader a clue, but far from the whole story.

It seems obvious to me that if one wanted a good answer to a data-analysis question, one would provide such basic information up front instead of waiting to be asked for it in a comment. But it's often missing. Perhaps the key reason is that people who are new to data analysis don't realize how much good data analysis depends on things like the goal of the analysis and the type of the data available. They have been misled by overly simplistic courses and textbooks that make it seem as if data analysis is just a matter of consulting a big book of formulae and rules, and there's little difference at the statistical level between predicting which horse will win a race and comparing explanatory models of political attitude propagation on Facebook. Aren't statistics just those weird numbers you put at the end of your results section that prove your research hypothesis is right?

It's true that this contextual information often isn't crucial. One could take a stab at answering many of these questions as-is. However, when the asker provides more specific information, one can provide much better answers. In particular, one has a much better chance of understanding what the asker really wants, if they phrased their question poorly—as novices often do, which is totally understandable. Sometimes the best response to a question is to focus not on what is explicitly asked, but on the asker's underlying point of confusion. Like a good teacher or consultant, we want to give askers the best guidance we can, not just answer the question literally.

Anyway, if you agree with me that it would be nice for askers to provide context, in this sense, what can we do to increase the rate at which askers provide context in their first write-up of the question?

  • 9
    $\begingroup$ I've sometimes pointed people to @PeterFlom's blog post, "How to ask a Statistics question". There's also What makes a good Cross Validated Question?. Picking or making up a model question might be an idea. $\endgroup$ Commented Jun 1, 2016 at 16:51
  • 10
    $\begingroup$ How can we get posters to do this is the key here; beyond doubt it is highly desirable. I think it's unfortunate that our Help Center is full of negatives whereas posts laying out models for emulation are mostly within Meta threads or elsewhere. But that's true across several fields: it's easy to list common specific errors in any language, but hard to explain how to write engagingly. $\endgroup$
    – Nick Cox
    Commented Jun 1, 2016 at 17:50


You must log in to answer this question.

Browse other questions tagged .