0

I have a study in which I have several promising variables, all concurring with previous literature and significant at alpha=0.05.

However, I have this one variable that is of special interest and this variable is significant even with the conservative bonferroni correction.

How do I go by slipping in my manuscript like "Hey by the way look at this guy, this guy is definitely not a coincidence"? Because I don't want to set a new alpha and lose all the other good and informative variables.

  • If you are unsure about basic research tools and how they are used in your field you must teach yourself properly how to use them. Asking this stackexchange is neither the right place nor sufficient. If you have a research supervisor ask them for help and guidance! If not, talk to experts in your field and/or take a course in statistics as it relates to your field. – user2705196 Jan 17 at 12:54
  • It's not a statistical question. – Paze Jan 17 at 14:49
1

First of all, I'd say this question is off-topic here, because it relates more to statistics than academia as a whole.

In any case the answer by Janosch is accurate, when doing a statistical test, an observation is either significant or not.

Sometimes people use stars to annotate the level of significance a particular result would be able to clear, for example:

  • * for p < 0.05
  • ** for p < 0.01
  • *** for p < 0.001

Here's another example from an R tutorial article:

enter image description here

|improve this answer|||||
  • Thank you, that seems to be what I was looking for. – Paze Jan 17 at 14:49
0

You dont. Either a p-value is below your alpha and than you consider it significant or if it is above your alpha level you don't. But there is no such thing as "more significant".

Regarding the Bonferroni Correction. If you do multiple testing than you need to correct your alpha values/p_values. It should not be based on whether you would lose other variables. Thats what you generally consider p-hacking.

But you can also try a different Correction Method. Usually you do not want to use Bonferroni because it has a low power. Many Statistical Software provide alternative. I would recommend the Benjamini Hochberg Correction

|improve this answer|||||

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.