2
$\begingroup$

There's a function from a statistical package whose printout I'm slightly confused about. Should I ask about it on SO or here? On the one hand, it doesn't have anything to do with coding/programming, but understanding what is printed out. On the other hand, it's not specifically a statistical question either.

To be specific, I'm talking about the prior_summary function of brms. I would like to know whether the lines without priors specified are the lines were an improper prior was used. The documentation seems to imply this, but I'm not completely sure. Thanks!

$\endgroup$
8
$\begingroup$

I think that interpreting the results of statistical analyses is unambiguously a statistical question. It's true that such results are basically always the output of some function from some statistical package, but that doesn't really change anything except the superficial aspects of the results. That is, the meaning of regression results in Stata output is the same as in R output, and the statistics presented will typically be very similar. I suspect that coming to understand how to interpret SAS's output should be sufficient, or nearly so, to interpret Minitab's output.

Note further that we have a number of open (and in some cases, highly upvoted) threads about interpreting statistical output:

Etc.

Therefore, I think questions about interpreting printouts from statistical packages are on-topic on CV.


Edit: The meaning of elements of statistical output that are specific to a package but not to the analysis is ambiguous. If the question is just a read-the-documentation issue, it might be off topic both here and on Stack Overflow. On the other hand, the boundaries for what is on topic listed in our help center includes a vague line about "needing statistical expertise". Your question could be about reading the documentation, but where they aren't understandable without statistical expertise, and so on topic in that way. (This would be analogous to questions like, 'I read such-and-such book, but I don't understand what they mean by this passage'.) I am arguing here that questions about interpreting statistical output are on topic, in general, but that doesn't guarantee that all questions about interpreting statistical output must be on topic.

$\endgroup$
  • $\begingroup$ The problem in my case is actually quite specific to the package though (i.e. doesn't really fit the criteria in your first paragraph), in that I'm asking for the meaning of blank lines. (Also it's not really the results of a statistical analysis; the function prints out the priors used by a model, not stuff like parameter estimates.) Do you still think it's appropriate to ask it here? Thanks! $\endgroup$ – WavesWashSands Jun 22 '18 at 19:04
  • 4
    $\begingroup$ I think it's appropriate. Please do take the trouble to explain what brms & prior_summary are & to quote any relevant passages from the manual - by doing so you're helping to make it a question of statistical interest. $\endgroup$ – Scortchi - Reinstate Monica Jun 22 '18 at 19:23
  • $\begingroup$ (+1) But stats.meta.stackexchange.com/a/4554/17230. You've provided good examples of how to ask about digital computer outputs. $\endgroup$ – Scortchi - Reinstate Monica Jun 22 '18 at 19:29
  • $\begingroup$ @Scortchi, I upvoted many of the answers on that thread, but not that one. While I usually agree w/ Nick, I don't there. I think that, in general, interpretation-of-output questions are absolutely on topic (as I argue here); that said, any otherwise on topic question could be closable by virtue of being unclear or too broad. I don't see anything specific to output interpretation questions as necessarily making them any more or less likely to be unclear or too broad than questions on any other subject. $\endgroup$ – gung - Reinstate Monica Jun 22 '18 at 20:26
  • $\begingroup$ @Scortchi Thanks! I will. $\endgroup$ – WavesWashSands Jun 22 '18 at 20:29
  • $\begingroup$ @WavesWashSands, that sounds ambiguous. If this is just a read-the-documentation question, it might be off topic both here & on SO. OTOH, the boundaries for what is on topic listed in our help center includes a vague line about "needing statistical expertise". Your question could be about reading the docs, but where they aren't understandable w/o stats expertise, & so on topic in that way. I firmly believe that questions about interpreting statistical output are on topic, in general, but that doesn't guarantee that all questions about interpreting statistical output must be on topic. $\endgroup$ – gung - Reinstate Monica Jun 22 '18 at 20:34
  • $\begingroup$ @gung: I have asked the question here - please take a look if it's okay. I do think that reading the docs requires some stats expertise. Thanks! $\endgroup$ – WavesWashSands Jun 22 '18 at 21:26
  • $\begingroup$ @gung: I took the moral of Nick's answer to be:-Don't just dump output into a question & ask how it should be interpreted. I think it's worth drawing attention to for the benefit of future readers of this post - all the same I entirely agree with what you've written. $\endgroup$ – Scortchi - Reinstate Monica Jun 23 '18 at 0:11
  • $\begingroup$ @WavesWashSands, that seems like a statistical question to me. $\endgroup$ – gung - Reinstate Monica Jun 23 '18 at 0:59
  • $\begingroup$ @Scortchi, I think your raising the issue of the potential inconsistency between these two answers is entirely appropriate. $\endgroup$ – gung - Reinstate Monica Jun 23 '18 at 1:00
  • 1
    $\begingroup$ I've edited my answer in that thread in an attempt to address what appears to be inconsistent advice. Unclear, ambiguous questions are poor, while well focused specific questions are good: that's the nub. $\endgroup$ – Nick Cox Jun 23 '18 at 7:40

You must log in to answer this question.

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