I guess it is a safe bet that the p-value will be alive and kicking in five years from now. It has been criticized for decades and there are definitely many researchers who are aware of its limitations but in a lot of applied disciplines the testing of hypotheses (or theories) is done without a thorough understanding of the underyling logic and the procedures are passed on from one generation to the next. I have elaborated a bit on this (and other problems) in a recent paper of mine:

Article Taking Feyerabend to the Next Level: On Linear Thinking, Ind...

Science continues to advance but old truths are still true. David Booth

I think it is simply a matter of ongoing education. Certainly, try and pass what I have learnt on to my clinical colleagues. I have managed to all but eliminate "statistical significance" statements from manuscripts I co-author and referee. P-values will still have a place, but it will (should) be much more limited than now.

ps - have a look at http://datamethods.org if you are interested in discussions about how to use and interpret p-values and other statistical metrics.

A couple of years ago the journal "Basic and Applied Social Psychology" banned NHST (null hypothesis significance testing):

https://www.nature.com/news/psychology-journal-bans-p-values-1.17001

I am not sure if this ban is still in place, but it is a strong signal.

P-values already in ICU what is only left is for the doctor to unplug the electric cables. They are already dead, use them at your own risk. They died with Fisher

Also contributed a bit on p-values follow the link Article The American Statistical Association (ASA) Statement of 2016...

You have asked two completely different questions. But as for your first question, the p-value is absolutely not going to disappear in five years. First of all, change in scientific practice does not happen that fast (e.g., Simmons, Nelson, and Simonsohn 2011--a very convincing and influential paper demonstrating why p-hacking is bad--was published 8 years ago and people are still p-hacking). Secondly, there is not even universal agreement that p-values are bad. (There is universal agreement that many people misuse p-values, but that doesn't mean that they themselves are not useful.) To see a perspective in favor of p-values, you can read opinions from statisticians or scientists who care about error control, e.g. Deborah Mayo or Daniel Lakens. see e.g. https://errorstatistics.com/2019/06/17/the-2019-asa-guide-to-p-values-and-statistical-significance-dont-say-what-you-dont-mean-some-recommendations

There are many excellent comments and interesting links in this thread. P-values are definitely not definitely bad "per se", but they are frequently misused. Here is another interesting paper which explains the origin of its current usage resulting from the dispute between Fisher and Neyman:

https://www.nature.com/news/scientific-method-statistical-errors-1.14700

Horst Treiblmaier We know today that opioid misuse is widespread. If I follow your logic would I not ban all opioids instead of trying to correct the misuse?. What about all the people that use these compounds correctly and have their pain relieved? It is, of course, possible that new drugs with different properties might replace opioids. Would we then ban opioids and leave the people that can't tolerate the new drugs in pain. In other words what % of the population has to misuse p-values before they are banned for everyone as BASP did with p-values? How does this make science better? D. Booth

David Eugene Booth Opioids are a great analogy!

I do not want to ban p-values completely, but just want to point out what they can do and what they cannot do. This knowlege might be common sense in some communities (e.g., statistics, psychology) but in many "applied" communities (e.g., business and related fields), methods are used blindly all too often. It resembles prescribing opioids for every ailment you can think of.

For example, quite often there is a dispute between qualitative and quantitative researchers. The former approach seems less rigorous since there are no numbers involved, but all too often the weaknesses of quantitative approaches are overlooked, since many researchers are simply not aware of them. This becomes especially problematic when they become reviewers.

In a nutshell, I personally do not want to ban the p-value but I think that such "bans" can help to raise awareness that limitations exist and it might motivate researchers to look at Bayesian statistics, just to give an example. The more methods we have in our toolbox, the better.

Horst Treiblmaier Thank you sir. First, I appreciate that you don't want to ban p-values. However, by looking in this thread we see that there are clearly those who do. 2nd, I am an accredited professional statistician (ASA) who spent 40 years teaching business statistics to undergrads as well as graduate research students. One of the things I taught was basic design of experiments. It is difficult to teach experimental designs when my department colleagues were praising quasi-experimental designs(it is OK if you don't randomize). Of course, we always find some that aren't interested in learning or miss the point no matter what you do. I was asked once by a PhD student It's really OK if I just take the first 100 people I come to for my sample right? However these methods are used for things like testing drug treatments for humans. He probably didn't grasp p-values either. The reason we use things like p-values is that they give us information about whether or not a particular hypothesis is true and/or under what conditions. If that is something I want to know I need the p-value information. Perhaps it could be phrased differently but I need the information. Bayes methods are quite interesting but even with present computers many important problems are intractable. Perhaps that too will change but the information I need to know will not change. I can hope for better phrasing but the needed information remains. I always felt that it was my job to help students understand what information they needed and how to get it. Sometimes I failed. That didn't change what they needed however. Perhaps I should have been better. Anyway. My point is one can ban p-values but you can't ban the needed information they convey. I apologize for the length. I thank everybody who is still reading. D. Booth

David Eugene Booth I personally do not want to ban anything, but I think it is an interesting idea that some journals do that just to give a signal that there are other options available. I believe that all empirical methods have their own strengths and weaknesses and the better we understand them, the better our research will be. The one thing I hate most about the p-value is that much too often it becomes the sole decision criterion and, for example, the effect size is completely ignored. Many researchers in the applied sciences cannot even explain how the p-value works, which produces many misunderstandings. Just pressing a couple of buttons in SPSS is simply not enough for good research. You also raise the important issue of experimental vs. quasi-experimental designs, which is also something I teach my PhD students. In the social sciences we frequently have to rely on quasi-experimental designs, since we do not have any other choice. This is fine, as long as one understands the limitations.

The decision is completely with the researcher- clinician’s hand.The p value is just a figure value to justifyfy the decision via mathematical fact.The significant of it reflects the sample size draw to proof the clinicians anticipation on effect size[ ES] of minimal clinical difference.The clinician - researcher are free hand to decide it and The decision is completely beyond the p value if the outcomes are clinically sounds enough to show it via the estimateed sample size.

Seyed Shahmy You are right, that is how it should be. In reality, the situation might be a bit different. Many researchers do not fully grasp the concept of the p-value and rejecting/corroborating hypotheses is a YES/No question. Recently I got a paper rejected (from a so-called top journal) since one out of three hypotheses was not significant (i.e., p value > .05). Food for thought...

Horst Treiblmaier , I was not your reviewer but I would not have rejected a paper that simply failed to reject one hypothesis. I would hope the reviewer did not do that and had other reasons that perhaps were not made clear and should have been.. I was in a similar situation recently with respect to a paper by someone else, that I tried to correct in a followup but was rejected by the first journal. I got a better hearing somewhere else. Given you experience I share this with you to indicate there are decent reviewers out there. I have attached that paper. please pardon my poor prose. I was awaiting open heart surgery as I wrote it. Best wishes, David Booth

David Eugene Booth All the best for your heart surgery!

It was not my intention to highlight the shortcomings of the peer review process (this is discussed in other threats), but just to give a practical example of how a flawed understanding of the p value might cause damage. Since authors are aware of that mindset, they frequently select only those results in which the p value is significant (p value hacking), which causes a huge bias in science.

There is an excellent paper from Ioannidis on that matter:

Article Why Most Published Research Findings Are False

If researchers/reviewers would better understand the p value (and its limitations) this would be highly beneficial for research in general.

P.S.: Thank you for sharing your paper.

This debate has been very helpful. I have accessed the suggested reading and its helping me a lot

The "Clinical significance vs Statistical significance" debate has been an ongoing one or several years now. Both are relevant and have their own places in the realm of medical research. I don't think that either of these is meant to undermine the role or importance of the other. Let us look at these independently as two separate measures or tools. In the absence of a distinctively supeior salternative to 'p' value, I suppose it will continue to have a place in medical research!

I believe both are important and each gives different messages.

Statistics deals with numbers and probabilities and thus gives a scientific validity to the results.Clincalsignificance emphasizes ondirect observationmade in clinical set up and thus gives information the usefulnessfor patient care.