Dear Dr. Arbind K. Choudhary , I do not think so.

I agree with the information provided by you while initiating this discussion, the statistical significance (what or how much is acceptable range? - is debatable) and its relevance need further investigation.

I completely agree with Dr. Anamitra Roy , the pandemic has compelled us do continue our research work without much of statistical tools.

It should be noted that the statistical test are one of the ingenious methods of extracting the hidden truth (pattern). It saves us time, effort and money.

Warm regards

I dont think as we use in agriculture

You must distinguigh the concept of statistical significance and its misinterpretation and abuse. The concept itself is reasonable and is one possible and legitimate safe-guard against over-optimistic interpretations of observed data.

To me it looks like that the "p-value" more and more serevs as a straw-man to be blamed for many much more relevant things that severly go wrong, like bad experimental designs, the lack of quantitative research (what are theoretically possible effect sizes? what are likely effect sizes given the data? what are the consequences of these effect sizes? what are predcitions by these effect sizes that can be verified or falsified experimentally? and where are such experiments actually done?), a missing culture of doing replication studies, an extremely low statistical literacy among empirical researchers. This list is surely not complete.

Dear Mubashir Ali Rather Thank you for your participation and contribution.

If the study is based on a very large sample size, relationships found to be statistically significant may not have much practical significance. Almost any null hypothesis can be rejected if the sample size is large enough.

a statistically significant treatment effect may exist but be biologically irrelevant because, although statistically significant, it is smaller than the predefined biologically relevant effect size.

While statistical significance relates to whether an effect exists, practical significance refers to the magnitude of the effect. However, no statistical test can tell you whether the effect is large enough to be important in your field of study. ... An effect of 4 points or less is too small to care about.

I would add to Jochen Wilhelm 's list: publication biases. It's no secret in the social sciences that most journals toss replication studies and still 'implicitly' look for p-values. Things are changing, albeit slowly. I'm seeing more and more journals emphasizing effect sizes, statistical power and CI than p-values in recent years.

Dear Jochen Wilhelm I appreciate your opinion. According to you, the concept of 'statistical significance' has been more often misinterpreted and abused. As such it is highly relevant. Thank you so much for your contribution to the discussion!

Dear Blaine Tomkins Faraed Salman Thank you each onefor your individual contribution to the discussion. I appreciate your individual views!

Arbind K. Choudhary

Dear Professor,

It is a very serious topic of discussion that you have raised. Many thanks to you for that.

A few days back, I was reviewing everything that I had done in 2020. I found that the researches pursued by me after the commencement of this pandemic were based on secondary data only. So naturally there was not much of a statistical analysis required. I talked to some of my friends and colleagues who are also in the same profession. I found that they were also of the same opinion. Not too many scientists and researchers were able to collect primary data during this period. So mostly researches were in the forms of protocols, methodologies, reviews etc.

To sum up, I think it is true that statistical significance test has been reduced in social experiments. I am not too sure about Biology though.

Best regards,

Anamitra.

Dear Anamitra Roy Thanks for your update and contribution. I appreciate your observation!

I am not an statistician, mathematician nor expert (sorry in advance), but I find it such an interesting topic. Most of the arguments seem directed towards Null-Hypothesis-Significance-Testing (NHST) and not p-values (see the title of [1]). Based on a "significant" p-value (< . 05) it is often concluded there is an "effect" or "difference". This is not what a p-value < .05 indicates nor what it implies. The p-value tells a story about the variability of the data. Using the term "significant" already implies NHST, so a cut-off is selected (p < .05), which seem not to exist [2]. Some seem to have argued to replace the term "significant" [3]. There is nothing wrong with using and displaying p-values, just as it is to use and display effect sizes. Switching to effect-sizes with confidence intervals contain the similar (if not the same), and more information [1,4]. I think mentioning both effect-sizes, with confidence intervals and p-values makes displaying results more transparent and comparing results easier. Yet, wouldn't the use of confidence intervals, bayes factors or any other statistics eventually lead to the same issue by selecting some cut-off value? I hope this somehow adds.

[1]researchgate.net/publication/270585309_Eight_common_but_false_objections_to_the_discontinuation_of_significance_testing_in_the_analysis_of_research_data/references

[2] Article The ASA's Statement on p-Values: Context, Process, and Purpose

[3] Article I can see clearly now: Reinterpreting statistical significance

[4] Article The earth is round (p

Dear @Wim Kaijser Interesting interpretation and explanations. I appreciate it. Thanks for your contribution!

The larger the sample size of the study, the more likely it can detect the biologically defined relevant effect as statistically significant. Statistical power increases with the sample size, while all other parameters of statistical testing remain the same.

Dear @Abdulhadi Mohamed Jumaa I consent to your views, and appreciate your contribution. Thank you 😊

Dear Dr. Arbind K. Choudhary , I do not think so.

I agree with the information provided by you while initiating this discussion, the statistical significance (what or how much is acceptable range? - is debatable) and its relevance need further investigation.

I completely agree with Dr. Anamitra Roy , the pandemic has compelled us do continue our research work without much of statistical tools.

It should be noted that the statistical test are one of the ingenious methods of extracting the hidden truth (pattern). It saves us time, effort and money.

Warm regards

I feel a bit out of place commenting (due to my position and experience), but a p-value does not indicate there is an effect/difference/causality based on the observed patterns and neither does it suggest they are "relevant." I can perfectly find an R2 of 0.01 and p < .001 based on any dataset I work with (e.g. y=ax+b=1.5*x+b). This does not say there "is" a positive effect/relation/causality between y~x and neither do I consider R2 0.01 "relevant." Assuming y~x is linear, my interpretation (which could be completely wrong) would be that the variability (probability) of the coefficient (a), based on the sample size and assumed model compared to the null model is < .001 (neither "significant" nor "nonsignificant"). If we assume the p-value is a relative measure of variability and sample size to substantiate I have taken "enough" samples (consider enough is a dichotomous statement as in "enough" or "not enough" similar to the NHST approach). As such, if before the analysis I expected R2 would be 0.6 and a > 6, but I obtain 0.01 and a = 1.5, to me it seems perfectly reasonable to suggest I did not observed a "relevant" pattern because p < .05. Hence, it is relatively difficult to describe patterns in a non biased (considering we know the bias?) fashion, since everyone has different expectations/knowledge of which the message is never completely communicated in an article. Hence, "significance" has lost its relevant meaning, but not the p-value, because the meaning of it was only relative (e.g. p < .05). Consider also playing around on this webpage (https://projects.fivethirtyeight.com/p-hacking/).