3

Recently got accepted for a poster presentation based on a pre-registered abstract. However, all the results are non-significant (after running the analysis) and I can't help but feel slightly disappointed. Has anyone faced a similar issue and presented non-significant results at the conference? :( How often does one see non-significant results at a conference? It sucks to think that this disappointment will be blown up to poster size for all to see.

1
  • Well, that just means that your null-hypothesis cannot be rejected with the exceptional set your chose. If your exceptional low probability set was a decent one, it merely means that it may be true after all and the alternative has not enough ground for validation. That is a perfectly valid outcome of a statistical study and you should be always ready for it. Whether it is worth reporting depends solely on whether the question was worth investigating in the first place. If it was, yes, the negative result should be made public and has equal value with a positive one.
    – fedja
    Jan 25, 2023 at 1:04

2 Answers 2

4

Publication bias is a huge problem in academic publishing right now.

In summary, if a field uses null-hypothesis significance testing with some arbitrary threshold (say, p<0.05) for significance, and only significant results are published because the field is only interested in significant results (impacting publishing decisions both by authors choosing what to submit and venues choosing what to publish), many (and possibly most) published significant results are actually showing effects that are not real or exaggerated sizes of the actual effects.

A high-quality study is high-quality whether or not the result is significant. You should design studies where either a significant or non-significant result is interesting (one key aspect of this is designing studies to be large enough that non-significant results are interesting).

Think about what is interesting about your non-significant result, and know that it is honorable and methodologically far better to present such results rather than hide them in a cabinet.

3
  • Evidence contrary to the hypothesis is just as valuable as any other.
    – Buffy
    Jan 23, 2023 at 18:32
  • @Buffy The biggest problem is with underpowered studies where no individual study provides useful evidence in any direction, yet where a sufficient number of such studies are performed such that those that spuriously provide evidence of a particular hypothesis get published as evidence of interesting large effects.
    – Bryan Krause
    Jan 23, 2023 at 18:35
  • "You should design studies where either a significant or non-significant result is interesting" - I can not stress it enough how important this step is for a research career. Almost an open secret of the academic world: for some reason, everyone seems to teach how to design a correct study, but almost never a good one.
    – Lodinn
    Jan 24, 2023 at 19:25
-1

If your method was promising but turned out not to work, explain both the promise and the failure in your poster.

Related: What to do when you spend several months working on an idea that fails in a masters thesis?

You must log in to answer this question.

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