Should this not be a must for any scientist and clinican?
The Following was posted by RetractionWatch, the web link is provided:
Recently, a biostatistician sent an open letter to editors of 10 major science journals, urging them to pay more attention to common statistical problems with papers. Specifically, Romain-Daniel Gosselin, Founder and CEO of Biotelligences, which trains researchers in biostatistics, counted how many of 10 recent papers in each of the 10 journals contained two common problems: omitting the sample size used in experiments, as well as the tests used as part of the statistical analyses. (Short answer: Too many.) Below, we have reproduced his letter.
"Dear journals: Clean up your act. Regards, Concerned Biostatistican
without comments:"
http://retractionwatch.com/2017/04/21/dear-journals-clean-act-regards-concerned-biostatistican/
I totally agree with you that articles that leave out such basics as sample sizes should not be accepted. However, some journals require, and thus perpetuate such errors as using Cohens d (a sample size dependent mockery of effect size) and the use of the classical (but seriously misleading) values of "small", "medium", and "large".
I completely agree and finally I'm really wondering why so many journals and their reviewer board (in case they have) are accepting this non-scientific (GSP = Good Scientific Practice) mode of behavior. Actually not acceptable or?
How often did I or one of my PhD students try to repeat something mentioned in an scientific article or a company's prospect (e.g. supplier of antibodies or assays or ...) etc. and the result wasn't the same or we weren't able to repeat because of a lack of information and I could really write a book about the bullshit answers when we contacted the other groups.
I know very well how it feels if one of your manuscripts will be rejected cause of mistakes or lack of clarity or missing information. But finally it is good and I'm always happy, that another scientific or clinical colleague investigated his valuable time to review precisely and correct and particularly in terms of mathematics (statistics)!!
The fundamental knowledge of correctly used statistics is something I'm missing really since my first step-in into the clinical work, I'm still wondering about. Unfortunately in life science disciplines it seems not to be better.
Dear community,
Have a nice weekend and think about it.
Kind regards,
Jochen
P.S. It is similar with graphs and all the lovely images, e.g. of cell culture work in context with cytocompatibility etc. Is all this the result of the huge pressure to perform?
I think that there are many methodology complications here, beyond simple valid statistics and written descriptions.
In social sciences, for example, studies often use elaborate statistics, but do not really account for ALL confounding variables, so despite pristine statistics, the findings cannot be easily reproduced.
My attached article about learning in the context of Complex Dynamic Systems is about language learning, but I think is relevant in many disciplines where there is not truly a single variable that can be fully isollated.
Article Environmental factors affecting Computer Assisted Language L...
Friends, thanks for the very valuable comments, which are highly appreciated.
I would argue that too many value calculated statistical results 'without critical thinking'.
For example:
'An apple found in a car is not synonymous of proof apples grow in cars' -
by this, there is a huge difference between correct observations vs wrong conclusion and interpretation.
Some may be interested in a very well students blog about
Statistical significance vs. clinical significance
http://www.students4bestevidence.net/statistical-significance-vs-clinical-significance/
http://www.students4bestevidence.net/statistical-significance-vs-clinical-significance/
just another example if I may
Mathematical mistakes makes results of Christian Tomasetti and Bert Vogelstein mean nothing.
"Choosing the distribution parameters so that the cumulative distribution function fits the incidence curve they get that only 13.5% of population is susceptible to colon and 22%-to prostate carcinomas. Unfortunately they use wrong formulas for Weibull distribution as well as for statistical criteria[5]and thus this conclusion is unfounded too. Nevertheless the main problem of this research is not the incorrect equalities but the hidden assumptions which, as it usually happens, are neither discussed nor even mentioned. "
……………
Medics usually tell us "Don’t treat yourself" and they are right as they have much more knowledge about diseases than we do. Maybe we may give them a similar advice: "Don’t do maths yourself", since we know it better?
https://www.researchgate.net/publication/310326046_Weibull_Distribution_Theory_of_Cancer_Age-Specific_Incidence_from_the_Mathematician%27s_Point_of_View
Article Weibull Distribution Theory of Cancer Age-Specific Incidence...
Björn L.D.M. Brücher, your "Don’t do maths yourself" certainly fits a huge number of researchers. However the real problem is that their research is essentially math based and they can barely do algebra. We need to increase the statistical (and hence math) training in any field where the researcher is going to be primarily measuring or counting. A lot of these unfortunates have been trained so that they are like a chemist who didn't get around to actually studying much chemistry.
Some of them unconsciously know their deficiency as is indicated by the fact that they often have a statistician in the store room or broom closit (really!) who is their supplemental brain and does not even get mentioned in the publications. Maybe most political scientists etc. should start with a bachelors in statistics with a minor in what they are going to go for in graduate school.
- Badges
- Science topic
- Similar topics
- Academic Journals