5

I know this will vary from journal to journal, but I'm working on my first publication and have had little guidance, so any personal experiences would be helpful to me.

When you present your p-values for highly significant data, where do you stop? For example, I have some p-values that are equal to 7 * 10^(-14) (or 7E-14, or 0.00000000000007). In this situation, where do you make the cut-off? Would you just say p < 0.00001? If so, an additional question arises - if I have two tests that are highly significant but are different from each other, say one has p = 2*10^(-10) and the other is p = 5*10^(-30), are they both reported as p < 0.0001 (or whatever the cut-off is)? Does it matter that one is so much smaller (and therefore more significant) than the other?

Glen_b
  • 257,508
  • 32
  • 553
  • 939
Peter
  • 53
  • 1
  • 3
  • I think this would be better suited to the statistics SE. Although the numeric difference between 2E-10 and 5E-30 is very large, I suspect that the *practical* difference between them is probably negligible. Note that the interpretation of p-values is rather subtle, and the level of significance should be chosen prior to the analysis (which essentially indicates roughly where the cut-off lies). – Dikran Marsupial Mar 12 '15 at 15:55
  • I think it is suitable here since it is a style question and not a stats question. – StrongBad Mar 12 '15 at 15:57
  • I considered the stats stackexchange board, but I guess my real question is getting more at the presentation/reporting of a p-value in a publication, not so much what is a p-value. I thought it would be more appropriate to ask people who publish frequently what they do in this situation rather than ask stats gurus if the values are important. I will delete and move the question if necessary, but I think the answer I'm looking for will be found here. – Peter Mar 12 '15 at 16:00
  • @StrongBad I quite disagree. In my opinion, it is not a good type of question for AC.SE. It's about "how do you write XY in a paper" where "XY" is specific to one branch. Only people from that branch can profit from it. – yo' Mar 12 '15 at 16:06
  • @yo' I am happy to be overruled. The branch of people who need to report p values is pretty big. – StrongBad Mar 12 '15 at 16:09
  • @StrongBad Well, I'm not sure. I cast the vote, but I'm open to discussion (that's also why I made the comment above). I also think it is a good question. And you certainly have a point that many people need it, but at the same time, the same is true for "presenting any statistical analysis", so should we accept such questions? – yo' Mar 12 '15 at 16:11
  • I'm still advocating for the question to be left open on AC (even though I understand the desire to move) because all academics who use statistics and publish those data in peer-reviewed journals will at one time have had to figure out the answer to this question themselves. I don't think it's restrictive to one branch of academia. I think this question would be relevant to academics working in almost any science. – Peter Mar 12 '15 at 16:12
  • Peter, welcome to AC.SE. The SE network can be scary and not welcoming at times especially when early experiences focus on moving your question. The whole idea of the SE network is to provide really good answer to really good questions. The different site attempt to what they consider on-topic and off-topic and this is a question that the answer is not clear. Have a look at our [help] and feel free to ask question in [chat]. – StrongBad Mar 12 '15 at 16:19
  • It is not necessary to delete a question and then asking it again some place else. You can just flag the question and as a moderator to move it. – StrongBad Mar 12 '15 at 16:22
  • 1
    I'm voting to close this question as off-topic because answers to this question hugely depends on the level of accuracy your research topic needs. Some need more and some need less. –  Mar 12 '15 at 17:53
  • @EnthusiasticStudent Isn't that part of the answer in itself? – Peter Mar 12 '15 at 18:03
  • 1
    See [here](http://stats.stackexchange.com/questions/78839/how-should-tiny-p-values-be-reported-and-why-does-r-put-a-minimum-on-2-22e-1), one answer of which already gives the APA-style answer and another answer discusses issues of meaningfulness of extremely small p-values and which is relevant to your question at the end. While it's not an exact duplicate (and so perhaps wouldn't close as one) I believe your questions are addressed there. – Glen_b Mar 12 '15 at 23:10
  • 1
    @Glen_b: +1 to your comment, but I think it *is* a duplicate and vote to close it as such. – amoeba Mar 13 '15 at 00:07

1 Answers1

3

For APA 6 Style

Decimal fractions—new guidelines for reporting of p values to two or three decimal places. (However, p values less than p<.001 should be reported as p<.001).

I believe in APA 5 Style you reported the exact p value when p is greater than your magic bar (e.g., .5) and p < .5 when p was less than the magic bar, unless it was less than .001 in which case you report p < .001.

StrongBad
  • 230
  • 1
  • 10