Not necessarily, statistically significant are the only results. Non-significant results are are also important results. It depends on the nature of the study, context, variables covered, statistical tools applied, etc.

Further, journals are not categorized on the basis of results significance levels.

Submit your paper to the relevant journal!

Not necessarily, statistically significant are the only results. Non-significant results are are also important results. It depends on the nature of the study, context, variables covered, statistical tools applied, etc.

Further, journals are not categorized on the basis of results significance levels.

Submit your paper to the relevant journal!

Dear Hussein,

I am afraid, you have not indicated the necessary information. Not even the scientific field. There are fields and subjects when statistical aspect is not necessary.  Regarding results, one should differ professional and statistical sides of outcomes. Often the statistical analysis has no sense. I think one can judge only using the case by case methodology.

I think, the young colleague should send the manuscript to an editor. He can get some useful experience.

Dear András.

I am not sure if I explained my point well, but I think that this information is sufficient to compile an overview of this issue in order to not be personal issue

Many thanks, Dr.Subhash  I told him that too

Thank you Dr.El Naschie for your prompt reply.

The fact that your results are not significant really calls for a lot to be considered. Either, your methodology, choice of variables, data transformation method, data timing among other things may need to be examined. It is of crucial importance to mention at this stage, what exactly you are looking at, so that Specialists in the area, have full information to advise you. What conclusions do you really draw from insignificant results???? Well, if everything is under control in the subject area, the insignificant results obviously means what they should identify.

I agree  with  you Dr.  Ahmed

Thank you Dr. Bonga Wellington Garikai , you're absolutely right

 It is important to take the results of some scientific research into account, whatever the outcome they are important and must be published if they are within the scientific contexts. Many of the research have been published earlier did not drive the attention of scientists at the time, but later became have scientific value after gathering information from other research.

Dear all,

As  Subhash C. Kundu says, important results could be no statistically significant. 

In new studies, they could indicate the tendency that one day in the future could  became statistically significant.

Have a nice day

Helena

Dear Colleagues,

Good Day,

This is a Reflexion from a paper attached in here, it says :  " Scientific journals in most empirical disciplines have regulations about how authors should report the precision of their estimates of model parameters and other model elements. Some journals that overlap fully or partly with the field of demography demand as a strict prerequisite for publication that a p-value, a confidence interval, or a standard deviation accompany any parameter estimate. I feel that this rule is sometimes applied in an overly mechanical manner. Standard deviations and p-values produced routinely by general purpose software are too often taken at face value and included without questioning, and features that have too high a p-value or too large a standard deviation are too easily disregarded as being without interest because they appear not to be statistically significant. In my opinion authors should be discouraged from adhering to this practice, and flexibility rather than rigidity should be encouraged in the reporting of statistical significance. 

If one study made a certain claim of statistical significance, and subsequent papers made citation to that paper without re-testing so that the whole field takes the same position as the first paper---but this young researcher did an empirical test and it shows that the original claim by the first paper is verified to be statistically insignificant. In this case, the finding of statistically insignificance by the young researcher is considered a new finding worthy of publication. It may point to the fact that the "old theory" is no longer valid or should be reconsidered.

Which journal? If the finding is a new contribution of knowledge to the field, most journals would welcome the paper.

Hi there,

there seem to be some confusion going around here. Do you have right sample size? (meaning if the sample size was actually calculated before the initiation of the study). Is the result is still negative? There is no problem here, the study still can be publishable! Because the negative result could still be significant  - it could disprove an existing hypothesis or theory, substantiate other negative findings, trying to address a controversial issue, etc.

If the sample size was not calculated, and because of that factor the result 'seem' in significant; this means, someone else has to repeat this work! This means, basically money wasted. 

Based on such work, you cannot initiate a treatment, care for a patient, you cannot create some policy or a plan to address the concerns of a particular population or a health care need. Because, what you reported might not be true, because of lack of sample size. right? If there was adequate sample size, it is possible that it might paint a different picture.

@ Mohamed: Scientists could fabricate and publish their results in Nature, Science, or a high profile journal. If someone finds out, the career will be over! The acceptance rate of high profile journals are 15-20%. Nature doesn't focus on any specialty, it can publish seismology, cosmology, exobiology, archaeology, anthropology, biology and medicine. You can imaging, how many papers that they can accommodate.

The editors and reviewers always evaluate to see if the results are worth publishing because of all these reasons - reliability, adaptability, applicability!

If you're publicly funded (meaning tax payers dollars), and if the results are insignificant (I am not talking about a negative results when sampling was appropriately done), what can the agency do with such results or how can they help their patients? 

Finally, why would someone wants to mislead others by publishing such inappropriate results or for that matter why would such finding should allowed to be published? This is one of the many reasons why the editors and reviewers reject the papers!

I don't know. I am sure that journals reject also papers with statistically significant results. Many time the reviewers and Editors find only pretexts

Dear All,

There must be some confusion here really. Participants have drafted very characteristic opinions on publishing a manuscript of which they  have practically no information except that the results are not statistically significant. I am afraid statistical significance is only one part of the evaluation. However, Hussein stressed that statistically significant results would be meaningful. This indicates that not even the author is satisfied with his/her results.  Such data depending on the subject and other merit can be published mainly in national journals.  

My personal experience is that many editors and reviewers evince little interest in non-significant results. Acceptance of null hypothesis means that treatments tested did not result in new discoveries in which case the old theories hold good. The researcher should consult a good statistician who can suggest diverse ways of analysing the data. Some times even if the main data are insignificant certain correlations/regressions may be significant. Many journals insist on novelty. If there is any novelty in the results this can be stressed. Finally it depends on the subject and the relevance and importance of the work.

I can appreciate and feel the pain of a researcher who happens to be in a situation as described by Hussein. I suppose what Hussein seems to be asking for is how to deal with such a situation. Here are some thoughts that may possibly help

1: Find out why were the results non-significant: Possible reasons could include (a) inappropriate experimental design, (b) insufficient number of (genuine) replications for the study to have high power (preferably 80%) to detect a prespecified (specified prior to conducting the study) difference between any two treatments as significant at say 5% level of significance, (c) Inappropriate statistical methods to analyse the data, and last but not the least,  (d) Incorrect use of appropriate statistical methods.

2: If the study had none of the above 4 issues, a statistically non-significant result is the correct result. Keep in mind that statistical significance does not necessarily mean the results are practically significant. Likewise, statistical non-significance does not necessarily mean the results are practically non-significant. 

Now that the study is over (and possibly cannot be repeated), how do you proceed? I suggest that you write your paper honestly accepting upfront the deficiencies in your study if there are any (such as in 1 above) and say that the results are suggestive / indicative that will need to be confirmed using a more powerful study. You can (alternatively) also highlight that this was an "exploratory" study (if indeed it was) to gain some preliminary understanding how the treatments etc might express their effects.     

Dear András Bozsik,

I agree with you that there may be confusion. I would say more than anything else that there is a "trivialization" of the problem. However, regardless of whether the results are good or not in a work, it is undeniable that the "industry of the journals" sometimes decides that the paper should be accepted on a discretionary basis.

Dear Dr. Subhash Chandra 

Thank you very much for your helpful answers.

In most cases, authors do benefit from comments from the the reviewers which is important for the specific paper as well as for future submissions. Even if the manuscript is rejected, there will be lessons to learn for improvement and subsequent publications.What would matter is how he argue his case; how he frame his "observed results."

The goal is to discuss the issue of the suffering of A young researchers and not flattening the case as Dr. Fabrizio   said

Hi'

It depends on the journals' discipline. Insignificant results are often significant. The medical journals, for instance, when correlating some ailment with a certain group of people, are sometimes satisfied that there is no relation between the two. Statistically this is insignificant  results (p>.10), but the lack of relation is very significant in this experiment. I think that medical journals could tolerate insignificant results.

Yes, Avishag Gordon,

You are very correct. Insignificant correlation or difference is also a significant  and important result.

But it depends on nature of the study.

Hussein -

Greetings.  -  "Statistical significance" is a highly misleading term, causing many to put emphasis on a p-value, when p-values in isolation are not meaningful.  You can change your conclusion by changing your sample size. This is quite problematic, and impractical.  If you can estimate a standard error, that is much more meaningful than going forward with a p-value.  You need a type II error probability or other sensitivity analysis with a p-value.  "Significance" is a term that most scientific studies should avoid.  One should look for what is a substantial result.  "Statistical significance" is jargon that is commonly misapplied, and confuses results. 

Please see the attached link to a letter to the editor of The American Statistician.  It was written in 1987, but from what I have seen on ResearchGate, I think many scientists currently realize that this is a problem.

 https://www.researchgate.net/publication/262971440_Practical_Interpretation_of_Hypothesis_Tests_-_letter_to_the_editor_-_TAS

Cheers - Jim 

Article Practical Interpretation of Hypothesis Tests - letter to the...

Dear Dr. James 

Thanks for the interesting post

In all results, some may be significant; some may not!

Dr. Hussein,

In the world, nothing is always equal. Sometimes or generally some processes or treatments are similar. In formal courses of statistics, students are induced to believe that the Normal distribution, the assumption of normal errors and additivity rules the world. This is not true as you know. In nature, some populations have normal distribution but one population which the member are classified does not obey such rules. In such cases, the researcher has to change your mind that is the best organic machine to interpret numbers, but with a huge difficult to multiply two numbers with three figures each one. Any table where the means are followed by a, ab, b, never the treatment with the letters ab must be used to make financial decisions, for example. The core of the problem relies on the dimension of the statistical field and in the fact the formal post-graduation courses offers similar lectures about statistics to different professionals in the same class room, and worst: anybody has to achieve the grade A, otherwise the student cannot be considered an “eminent researcher”. In my opinion, research companies must have one department of statistics where the process under development has to be discussed with statistician professionals. In agronomy, I have seen international companies paying a lot of money to scholars for studying the application of agrochemical in some specific situation, but I never heard that the data set is sent to their Statistical Department to be analyzed and thus returned to the agronomist for his last consideration. I agree with the consideration from the Dr. James Knaub. Furthermore, I said before: the first decision is to make an EDA (exploratory data analysis) and next to make the decision of consulting a statistician professional. Scientific or development projects must have to anticipate grants to pay for such services. In statistics there is a rule: no ring is stronger than its weaker link. I hope these considerations can help question followers that can access the website of DATAPLOT for more information and concepts.

Best regards.

Walter

Dear Dr.José Walter Pedroza Carneiro 

Thanks for the interesting post,I agree with you

It depends on objectives and hypotheses of the study, even non significant results may satisfy your objectives and test your hypotheses.

Good point Dr.Subhash C. Kundu

Thanks Dear Saeed Noori 

Significance? A good explanation in the link.

Other issues are studies with negative results (e.g. some clinical trials). They are not easily published (for many reasons); however, science will lose useful information.