[ad_1]
PICTURE: Image of the computer with statistics on the screen.
view more
Credit: Pixabay
Scientists should stop using the term "statistically significant" in their research, recommend this editorial in a special issue of The American statistician published today.
The question, the statistical inference in the 21st century: a world beyond
This special issue, which contains 43 articles by statisticians from around the world, should help rethink statistical inference by starting a process that will advance statistical science – and science itself – into a new era.
Dr. Ronald Wbaderstein, Executive Director of ASA, Dr. Allen Schirm, retired from Mathematica Policy Research, has retired. Professor Nicole Lazar, of the University of Georgia, said: "After our review of articles in this special issue and in the wider literature, we conclude that it is time to stop Use the term "statistically significant".
"No p-value can reveal the plausibility, presence, truth or importance of an badociation or effect. Therefore, a label of statistical significance does not mean or mean". does not imply that an badociation or effect is highly probable, real, true or important A label of statistical non-significance does not lead to an badociation or effect that is improbable, absent, false or unimportant.
"Therefore, for the integrity of the scientific publication and the dissemination of research, the fact that a p-value exceeds an arbitrary threshold must absolutely not be taken into account when choosing the results to be presented. or to highlight. "
The articles in the special issue suggest alternatives and complements to predictive values, and highlight the need for widespread reform of editorial, educational and institutional practices. [quotes below].
Although there is no single solution to replace the disproportionate role that statistical significance has now played in science, there are solid principles for the use of statistics, say authors of the report. # 39; editorial.
"The statistical community has not yet converged on a simple paradigm of using statistical inference in scientific research – and might even never do it," they acknowledged. "A unique approach to statistical inference is an inappropriate expectation, rather we recommend to scientists performing statistical badysis of their results to adopt what we call the ATOM model: Accept uncertainty, be think, be open, be modest. "
This special issue of ASA builds on the highly influential ASA P-Statements and Statistical Significance, which has recorded more than 293,000 downloads and 1,700 citations, an average of more than 10 per week since its publication in 2016 .
AUTHOR'S QUOTES
Need for change
"Considerable social change is needed in academic institutions, journals, funding agencies and regulatory agencies, and we suggest working with science reform movements and reformers in disciplines, journals, funding and regulators to promote and reward a "replicable" science and reduce the impact of statistical significance on publication, funding and promotion ". – Good man
"The evaluation of manuscripts intended for publication should be" blind to results. "In other words, the aptitude of manuscripts to be published should be evaluated according to the importance of the research, without taking into account account of reported results. " – Locascio
"Everything should be published in some form if everything we measured made sense before we got the data, because they were potentially linked to a research question." Journal editors should be proud of their comprehensive sections on methods and base their decisions on their relevance. " a study to be published on the quality of its materials and methods rather than on its results and conclusions; the quality of the presentation of the latter should be judged only if it is established that the study is useful on the basis of the materials and methods used. "- Amrhein et al.
"Reproducing the research should be encouraged by giving line status to researchers who are reproducing studies." We would like to see the digital versions of the documents dynamically updated to display "Reproduced by …" under the names of the authors. authors of the original research or "Not yet reproduced" until it is reproduced. "- Hubbard and Carriquiry
"An important role for statistics in research is the summary and accumulation of information.If the repetitions do not give the same results, it is not necessarily a crisis." but the natural element of the evolving process of science The goal of scientific methodology should be to direct this evolution towards ever more precise descriptions of the world and its functioning, and not towards an ever-increasing publication. development of inferences, conclusions or decisions. "- Amrhein et al.
Alternatives and complements to p-values
"A number of factors should no longer be subordinated to"
"Terms such as" significance "badociated with p-values and" confidence "in interval estimates mislead users into overconfident claims We suggest that p-values measure compatibility between hypotheses and the data and interpret the interval estimates as "compatibility intervals" rather than "confidence intervals". – Amrhein et al.
"Continuous p-values should only be used in conjunction with the" false positive risk "(FPR), which answers the following question: if you observe a" significant "p value after a single unbiased experiment, what is the probability that your result is a false positive? "- Colquhoun
###
About the Taylor & Francis group
The Taylor & Francis group partners with researchers, learned societies, universities and libraries around the world to bring knowledge to life. As one of the world's leading publishers of scholarly journals, books, ebooks and reference books, our content covers all areas of the humanities, social sciences, humanities and social sciences. behavior, science, technology and medicine.
From our network of offices in Oxford, New York, Philadelphia, Boca Raton, Boston, Melbourne, Singapore, Beijing, Tokyo, Stockholm, New Delhi and Cape Town, Taylor & Francis staff provide local expertise and support to our editors, societies and authors. and a responsive and efficient customer service to our colleagues in the library.
Source link
