6
$\begingroup$

I am trying to do a community talk about hypothesis testing and want to focus on significance testing and p-values in particular. They have been a bit in the news as well recently (with some journal banning their use) and I wanted to try and present a balanced viewpoint on what they show and what they don't, reporting effect sizes, alternatives etc. So basically I do not have a specific question but wanted some hints on where to start reading about this to formulate the understanding and material for an informal talk.

So what I was wondering if I asked the CV community about some must read online and offline resources on this topic, would that be a valid question given that it is quite broad? I would guess this is not an acceptable question but wanted to confirm.

| |
$\endgroup$
  • 3
    $\begingroup$ There are two dangers -- (i) breadth and (ii) inviting answers that are purely opinion-based. I think you might be able to make one, or perhaps more than one, question that would avoid those dangers, and then perhaps raise the rest on chat (chat.stackexchange.com/rooms/18/ten-fold), where opinion-based and broad questions could be discussed (though with less chance of attracting a lot of attention) $\endgroup$ – Glen_b Mar 11 '15 at 3:20
  • 2
    $\begingroup$ @Glen_b, interesting to see that we were thinking the same things. $\endgroup$ – gung - Reinstate Monica Mar 11 '15 at 3:28
10
$\begingroup$

This doesn't seem like a bad question, or that it should be off-topic necessarily, to me. Questions asking for , and questions asking about the theoretical / philosophical aspects of hypothesis testing have long been on-topic. Instead, I see two potential issues:

  1. Try to frame your question in such a way that there can be a right answer or an answer that can justifiably be selected as the accepted one. It may be worth reading the classic SE blog post Real questions have answers.

  2. Be careful that your question doesn't become too diffuse. You are asking about a lot of topics ("what they show and what they don't, reporting effect sizes, alternatives etc."). That isn't necessarily out of bounds—it depends on how well you can weave them together. If necessary, you can ask a series of focused questions, where each in turn ties in to the threads that preceded it (by linking, etc.). Also be aware that some of those topics have already been discussed on the site, so be sure you've searched thoroughly first.

| |
$\endgroup$

You must log in to answer this question.

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