4
$\begingroup$

A recurrent issue in our forum is applied analysis questions. Not only are there ample examples in the CV network, but it comes up in Meta repeatedly. I think questions of the nature are appropriate for this network, but I would say less than 1% of these questions are detailed enough to get appropriate answers. It reflects poorly on the community. How can CV improve this situation?

I propose we create an tag to guide users on what constitutes adequate information to get high quality answers. If users are not able to conform the question to these standards, it should be closed for being "too broad"-- since it forces us to answer in an abstract, turnkey manner that is too nebulous to be useful (if we can answer at all). This is exactly the approach that we take with which is similarly problematic. With self-study, our site policies have improved our answers and questions dramatically.

There is no existing analysis tag.

I propose that the analysis tag be written in the following fashion:

Analysis: For questions dealing with the applied analysis of a specific dataset or design of experiment. Posters may be either statisticians or non experts in the field. Questions need to address the following issues:

  1. The field of application (e.g. biostatistics, econometrics, social sciences...)
  2. A basic statement of hypothesis or hypotheses
  3. A detailed description of the dataset including the sampling method used, measures collected, the measuring methodology (e.g. mass spectrometry, questionnaire, physical exam)
  4. Basic proposed data analysis plan
  5. Precise description of the problems encountered

If others concur, I am happy to resurrect some unanswered "analysis" questions and post in a comment that the user should consult the "analysis" tag to improve the quality of the post. If it is not addressed in a week, we can close it.

As a statistician, working with other scientists to discover what they want to do is my favorite part of the job. At work I encourage others to follow this discipline. What might we add to this tag to make it better applicable in this circumstance and invite more non-statistical researchers to participate in the forum?

$\endgroup$
  • 5
    $\begingroup$ In addition be being perplexed by "flag" like @AndreSilva, I should note that the [self-study] tag is a meta tag (which are supposed to be verboten) & hence is a little bit of a necessary evil. Of course, that doesn't mean that we shouldn't also have [analysis] as (another) necessary evil. $\endgroup$ – gung - Reinstate Monica Jul 11 '15 at 15:24
5
$\begingroup$

I came across the tag and was going to write a Meta post suggesting deleting this tag on the grounds that it is a meta tag (the only meta tag that we are tolerating here on CV is [self-study]) that moreover is inconsistently used and does not seem to serve any useful purpose. One could in principle put [analysis] tag, as it is defined in its wiki excerpt (see below), on at least half of the posts on CV. It's way too broad.

Before writing this post I searched Meta for "analysis" and found this question. It left me puzzled. The tag was suggested by @AdamO in July 2015. The post got only one comment and it was rather critical. There were no answers. The post did get three upvotes and no downvotes, but three upvotes seem to me to be only a meagre support. However, @AdamO went on to create the tag, wrote a detailed wiki, re-tagged a whole bunch of older questions, and by now there are 171 threads in this tag.

Was there perhaps some discussion elsewhere that I cannot find?

The wiki excerpt says:

For questions dealing with the applied analysis of a specific dataset or design of experiment. Posts tagged analysis are requesting statistical consulting assistance from the network. Questions tagged analysis need to be phrased appropriately.

We have tens of thousands of questions "requesting statistical consulting assistance", not 171. The tag is an obvious meta tag, and I don't see any benefit from it. People use this tag whenever they are writing a question about any "analysis", be it principal component "analysis", partial least squares "analysis", regression "analysis", etc.

I downvoted the question and suggest to burninate this tag.

$\endgroup$
  • 1
    $\begingroup$ (+1) The advice given is great, but a tag doesn't seem the right place for it. $\endgroup$ – Scortchi - Reinstate Monica Jan 23 '17 at 14:53
  • 1
    $\begingroup$ I'm mixed on this. The idea that we need to have some guidance for askers seems right, whether we want that in a tag wiki or as a meta.CV thread eg is ambiguous. A tag might be useful, but I think you're right that "analysis" is probably not a good name for the tag, if we go w/ a tag as the solution. Applied-assistance might be something people are less likely to screw up. But I wonder if a special meta.CV thread, like How to ask a “good” question on CrossValidated?, which we link to from the FAQ, might be a better strategy. $\endgroup$ – gung - Reinstate Monica Jan 23 '17 at 16:02
  • 1
    $\begingroup$ I am fine with deleting it, despite my valiant, er, nagging effort. As we all know, there is a problem on this forum of questions going unanswered. Reviewing many of these, the theme I found was that it often seemed to be applied researchers asking questions that were not statistical (or if they were, it was hard to suss out the actual question). The community has recently upped its effort to close such questions as off-topic which seems to be a better approach, although a bit harsh to those who are interested but untrained. Nonetheless, this isn't a consulting service. Fine by me to delete. $\endgroup$ – AdamO Jan 23 '17 at 16:10
  • $\begingroup$ @AdamO I did not want to come across as too negative; sorry. I agree with Scortchi that the tag wiki that you wrote for [analysis] contains excellent advice; perhaps we can indeed move it somewhere to Meta as gung suggested? Maybe indeed to that How to ask a good question thread (post another answer there?), even though I have to say that I find that thread too noisy to be very useful (the Q is very long, lots of comments, answers are too long as well). Regarding the tag, looking at it now after 1.5 years, do you think you can say that it helped? $\endgroup$ – amoeba says Reinstate Monica Jan 23 '17 at 20:47
  • 1
    $\begingroup$ I'm thinking a meta.CV thread would be better than a tag. I wasn't literally thinking we should add to How to ask a good question, though, that was just an example. I was thinking we'd have another thread dedicated to this issue. It could have 1 answer discussing how to ask such questions, & another answer discussing how to answer such questions. I'm not completely opposed to the tag idea, but I think meta.CV might be better. $\endgroup$ – gung - Reinstate Monica Jan 23 '17 at 20:59
  • 2
    $\begingroup$ @amoeba you weren't overly pessimistic, and I echo whatever pessimism is there :) I can say in confidence that my tag did not help. Most of the applied researchers who come by way of this site do not tend to become active participants in any sense: they do not upvote or mark answers as correct, provide edits to their own question, or ask further questions. I think we should just accept it as a necessary evil of running such a site. $\endgroup$ – AdamO Jan 23 '17 at 21:23

You must log in to answer this question.

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