Dear Mariano, 

We have some researchers in the RG that do lots of papers in statistics that are apllied to medicine or other health science.

Statistics is an objective  tool that let us analyse whatever subject we need.

Helena

Renzo, I agree with the first part of your answer, but do not agree with the second part. There are statistical methods which are objective and their properties are secure (Objective) and others whose properties depend of the verification of certain hypothesis and these are uncheckable (Subjective).

On the added third part, not all reality is in our mind and you can see explained this in the article of Javier Olivera I have put after. You can think that in the closed door there is not a door, and you could try to entry because you could think that there is not a door, but the closed door will say the objective reality of its existence. The door is not a product of our definition or of our mind alone, it is an external reality. Not all is theory.

Dear Helena, I think Statistics is an instrument to clarify a conjecture about medical research, but the main part of statistical analysis or applied statistics in Medicine are subjective. For this my question, is an objective Statistics possible in Medicine?

Interesting question.  The inanimateness of statistics make them very objective as Salvoatore said.  It is the subjectivity/objectivity of the person running the statistics that makes them objective or subjective.  As Ronald Coase said (paraphrased), you can use statistics to torture the data until it confesses to anything, which is in line with the Mark Twain quote above.  I am sure medical research, like many other fields, has some (likely few and far between) biased researchers and improper analyses, but there is no way to know which are objective and which are subjective.  That is why replication in science is so important.  The more replication, the more solid the support. 

It tests the objectivity of reality.

Eric, you can see a way in:

Ruiz Espejo, Mariano (2014). Fundamentos de la Inferencia Estadística Objetiva [Foundations of Objective Statistical Inference], 3rd edn. Lulu Press. Raleigh, NC.

http://www.lulu.com/shop/mariano-ruiz-espejo/fundamentos-de-la-inferencia-estad%C3%ADstica-objetiva/paperback/product-21821572.html

Juehui, the reality is objective, and Statistics describes, or infers from a sample, such objective reality of the population or a characteristic of it.

Let me rephrase it. Though reality is as is, we perceive it subjectively in our own lenses. Statistics therefore is a reliable and powerful tool to help us to better understand science objectively, if used correctly.

Medical statistics and biostatistics have been widely recognised as  recognized branches of statistics.  Medical statistics also involves the usual methods of summarizing, collecting, presenting and interpreting data in relevant areas of medical practices, These techniques are commonly and effectively used for estimating the magnitude of associations and test hypotheses.

Relevant advancements in this area can be found in the journals like Statistics in Medicine (Wiley, USA).

Statistics, in general, and Bio-informatics/ Biostatistics in particular, has long been applied to test/ verify/ validate research propositions scientifically/ empirically. Objectives are mathematically (and statistically) programmed and solved (as part of programming (linear & non-linear). Application of Statistics in 'Medical Research' is wide and immense; it is getting new dimension every day. We can still expect much wider & varied applications, in the days to come. Nice point to debate and cite examples of applications.

Yes, statistics can be used as a tool for data analysis in medicine, However, there are a lot of precautions of direct application of statistics to medicine/biology. Everything is dependent from the your experimental design, the size of the population etc.

Dear Sir!

it cannot be an objective but it is  a way to achieve your objectives!!

Haftamu, I think that for having a correct Medical Science it is necessary good means (as Statistics) and good ends (Ethical and Moral actuations as objectives).

@Espejo:

Statistics is not only being used in medical research..... It can be applied to many fields; Engineering, Accounting, Insurance, Banking, Psychology, e.t.c. It helps to make inference on the analysis performed on a collected data.

Dear All,

Statistics is a powerful instrument to be oriented in scientific results and their interpretation. The problem is the use of not suitable methods and the lack of proper data. Unfortunately, the teaching and the practice of statistics are not appropriate in many universities.

I suggest to look at this URL and the paper

http://metrics.stanford.edu/

http://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.0020124

Maybe it can be an approximate measure because no one can be certain of what happens on individual basis.

Regards

Carmen Camilleri

Dear Scholars: the Question As I understand is not what is statistics?but can it an Objective for medical Studies?

I think all the tools, including Accurate  statistical methods,Measures,and inferential stat,can be part of the Objectives, to give clear picture

of your investigation in medical sciences.

Dear Sir:Perfectly your comment is true!!

Dear Mariano, when people are diagnosed with a disease, they surely want to know their chances of survival.  Recently, a friend was diagnosed with a benign growth inside his ear.  We all wanted to know if it could be removed, and how it would affect the quality of his life if it could not be removed.  We wanted to know how fast this tumor would grow if it could not be removed.  What was the chance that his hearing and balance would not be impaired by this growth?  It's all about statistics, simple statistics.  I attach this link.  I quote:

'Statistics are estimates that describe trends in large numbers of people. Statistics cannot be used to predict what will actually happen to an individual.

Survival statistics for different cancer types, stages of cancer, age groups, or time periods can vary dramatically. People are encouraged to ask their doctor for the most appropriate statistics based on their individual medical condition.'

The pic shows some stats for colon cancer survival that I find interesting.  Thanks.

http://www.cancer.net/navigating-cancer-care/cancer-basics/understanding-statistics-used-guide-prognosis-and-evaluate-treatment

About the objective things, I recommend this link (in Spanish).

Renzo, in this article Javier Olivera explains that objectivity does not depend of the mind of the subject who observes. He convinces.

http://infocatolica.com/blog/notelacuenten.php/1506150613-icomo-se-demuestra-la-existen#more28616

I would like to ask, why not? Let us imagine what would happen when we went to a doctor, the doctor would ask us several questions, and based on those, giving us a diagnosis, or an "educated guess". Could we really say that is objective?

The decision of doctors are based on experience and instinct. However, in terms of experience, no human beings can do better than an statistical based algorithm with well-collected dataset. With this kind of algorithms, we could even know how much the probability that someone might be in danger of lung cancer, for example, because of her/ his smoking habit, how much is due to her/ his job or life-style or if her/ his partner, parents smoke. These details are all unachievable without statistics. Let us say so, without statistics, how could we identify if you, or your partner quit smoking would benefit the both more?

I would like to ask, if statistics was still not objective enough, how could we, as human beings do better? Please see the attached links for evidence.

Sincerely, Kuan-Wei

http://www.edventure-ga.com/apsistat/Activities/Articles/Making%20Health%20Care%20Better.pdf

http://allthingsd.com/20110912/wellpoints-new-hire-what-is-watson/

Thanks for pointing out this question, Kuan-Wei.  This is an interesting thread.

My first thought on this question was how it has occurred to me in the past that many statistical/medical studies, as well as sociological ones, may often be very tenuous, when there may be very many variables involved.  So when I saw where Eric noted the need for reproducible results, this sounded related. This could mean replicating experiments, where there may have been an error, but it could also mean what occurred to me: that sometimes there are so many possible variables involved that they may not all be known, and those that are known may be emphasized in different ways such that two different researchers, using the same data, may each legitimately arrive at substantially different conclusions.

Statistics cannot show causality, only correlations.  A scientific study in any discipline should start with a reasonable theory, based on the subject matter.  But all scientific models should be tested to see if the data may or may not support them. George Box said all models are wrong, but some are useful.  Einstein is credited with saying a model should be no more complex than necessary. I believe that is because a more complex model is going to depend too strongly on a given test data set, and that data may not represent the entire universe well enough to do that.  In physical models, changes are made to the model as more is learned. A very good example is the development of models for atomic structure.

So my point is that although statistics can be very helpful, and no scientific study is complete without it, i think that medical statistical studies may often be of such complexity that there may be many factors involved that are not obvious.  Data may be faulty, theory may omit important factors, or include some that aren't important, and the types of analyses can lead to different conclusions. Should we give up?  No.  Experiments need to be replicated. Studies need to be replicated. Sensitivity analyses are important. Try different statistical approaches.  I suppose some medical studies may be rather straightforward,  but I think there may be many cases where objectivity very much requires keeping an open mind to various possibilities. 

One might think that a result arrived at without intentional bias in the analytical procedures will supply THE answer, which would be very objective, but statistical studies always involve uncertainities, and even uncertainties about the uncertainties, so being 'objective' here would really mean being open to various possibilities and to verify, re-verify, validate, and re-validate.

To be 'objective' in a further sense, metadata are important.  These are data on/about the data. It is also important to document the methodology used, including statistical methodologies.  This will facilitate understanding the results of a study, and comparison to other studies.  

Cheers. 

James, causality implies correlations, and these are put in light by Statistics.

The point I was making was that you cannot assume causality from correlation.  

Yes, James, and if we obtain correlation one in several samples, there is a very high probability of causality.

Barbara, if applied Statistics in Medicine is subjective, a goal of a good research would be Objective Statistics, also in Medicine.

Dear Mariano, on the one hand, Statistics is an objective science if you use it correctly. I have read your papers and books and I know you are an expert on field and you use Objetive Statistics.

On the other hand, you must be very careful with the use of it in other areas and even more in Social behavior or medicine since we can extrapolate results only in very special cases.

As a conclussion, it depends on the researcher and the techniques he/she uses but if you use them correctly I think you can state that it is an objective science.

Dear all,

in my opinion, right now our problem is neither statistical methods nor technologies to build good algorithms, but to collect a dataset "well". I believe that statistics itself is already mature enough to deal such questions. Many biases or errors occurred not because of Statistics, rather of a bad dataset. 

Luckily, in medical studies, we could do actual experiments, which means the datasets still would have room to be improved, not like myself for example, in Economics, facing a "dirty" dataset is our daily routing, and I think that is why there came a new discipline "Econometrics".

In medical, or any natural science that we could do experiments repeatedly, such that we could obtain a "clean" dataset, I am optimistic about our progress. I believe that as long as we could have a correct dataset, in the very near future, we do not even need statisticians, or doctors, in this discuss, to express the  conclusions. That would be the works of algorithms. That might sounds a little radical, but I think we are already standing at the very edge of creating something close to "artificial intelligence", although almost every scientist tries to avoid this word in order to avoid ethic issues.

Sincerely, Kuan-Wei

http://www.wired.com/2009/04/newtonai/

Kuan-Wei, certainly in Medicine it must be respected the autonomy of the patients and his/her will when it is possible. Moreover the data of the patients are confidential and they are not open to all the researchers. Similarly, in clinical research the sample in which is studied a new therapy is reducted to a sample of volunteers. For this reason the conclusions of a study is limited to the volunteers, not to all the patients nor all people. Mathematically this actuation implies biases in the estimations which could not be corrected without the participation of a sample of nonvolunteers. But is this ethical in any case? It is not ethical in much cases.

Other key issues are the objective unbiasedness of the estimator and the objective unbiasedness of the variance of the estimator.

Dear Espejo,

Objectivity is generally considered to be an epistemic virtue: it marks an

inference as unbiased and trustworthy and grounds the authority of science in society. For example, medical drugs will not be admitted to the market unless there is objective evidence that proves their efficacy.

The objectivity of statistical inference is challenged by (i) the existence of several well-founded and competing paradigms for making such inferences; (ii) the partial reliance on subjective factors in these

inferences.

Sprenger (2014) established:

1. Better evaluation of statistical data. Improving the logic of NHST (→ Subproject A) solves several problems of great practical interest, such as the appraisal of insignificant findings and reconciling Bayesian and frequentist analysis. These results contribute to fighting publication bias

in science and close a salient gap that current standard methods (such as p-values or confidence intervals) leave open. Hence, almost all empirically working scientific disciplines can benefit from

our innovations regarding the interpretation of NHST .

On the policy side, expert findings such as the IPCC reports on climate change, are often criticized as not living up to their objectivity claims. However, these critiques are often based on outdated and

misleading ideas about scientific objectivity. A more refined account of objectivity in statistical inference will help to respond to these critics, assist the writers of such reports in formulating their

conclusions with the appropriate care, and support the authority of science in the public arena.

2. Better understanding of human reasoning. Statistical, causal and explanatory inferences are all cornerstones of human reasoning, and integrating them is one of the big challenges for experimental

psychology. This project contributes, especially in Subproject B and C, to the theoretical foundations for such an integration. At the same time, it investigates empirically to what extent such inferences have an intersubjective basis.

3. Fighting bias in evaluating medical trials, and better decisions in clinical care. Our research analyzes several forms of bias in clinical trials and shows that some of them need not be a problem for the epistemic authority of medical research (e.g., sampling bias). We also show how modifying the logic of statistical inference in medicine, and the interpretational perspective on those trials (→Subproject D), can lead to better appraisal of medical evidence and superior decisions in clinical care.

Thank you

Adeleke IT

020936H

Ibrahim - 

I agree that NHST (a lone p-value) is very problematic. (Note the letter to The American Statistician found through the attached link.) 

Cheers - Jim 

PS - Could you provide the full reference for Sprenger(2014)?

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

@Mariano; corrections on the nomenclature, taken.

@James;

Kindly find the full reference below as requested:

Sprenger J. Making Scientific Inferences More Objective. ERC Starting Grant 2014. Available at: www.laeuferpaar.de/SynopsisERCProject.pdf. Accessed on June 19, 2015.

Thank you,

Adeleke IT

040936H

Statistics is objective depending on the data used for the research

Dear Mariano,

It depends on the levels of evidence (i.e. high-quality randomized controlled trial, high-quality meta-analysis, cohorts or case-control study). We all are looking for certainties, but in medicine things are more complex than appears because of interindividual variability. It is about correlation vs. causation, independent vs. dependent variables. Questioning is more critical in preventive medicine, were taking action is often expected instead of waiting for the results of extensive research.

For example, a current issue is the anti-vaccination movement due to some secondary outcomes/associations reported, therefore more and more children remained unimmunized although this kind of immunization has proven effectiveness over time. And gaining control over epidemic consequences would be a very difficult task in the context of global antibiotic resistance… a pessimistic scenario!

Statistics per se is an objective and indispensable instrument in medical research, but results should be interpreted very carefully.

Oluwafemi and Renzo, yes, the objectiveness of Statistics can depend on the truth of the data, on the sincerity and correction of the user, and also on the used statistical techniques. With all correct, it is all right.

As Ronald Coase said, “If you torture the data long enough, Nature will confess.”

Other interesting book on this subject is:

Ruiz Espejo, Mariano (2014). Investigación Ética y Bioestadística [Ethical Research and Biostatistics]. Lulu Press. Raleigh, NC.

http://www.lulu.com/shop/mariano-ruiz-espejo/investigaci%C3%B3n-%C3%A9tica-y-bioestad%C3%ADstica/paperback/product-21822728.html

Yes, Renzo, and also Statistics can serve to adquire knowledge approximating the truth.

Answering this question, I think that while clinical volunteers have freedom to participate or not in a biomedical research in any time, it would be impossible to assure an objective statistical inference or testing. Other solutions can come from economic and assistential compromises without bad practices, but not only from Statistics.

Yes, Statistics would be an objective measure in Medicine, but it would be true that all people should be volunteer. Otherwise, Statistics would be subjective as it is now in this time.

Yes,it is

Statistics as mathematical science can or could be objective. But experimentation with human beings is not mathematics, it has to see more with moral and ethics.

A near publication of objective statistics and its possible use in observational medicine is this book:

Ruiz Espejo, Mariano (2017). Ciencia del Muestreo (Spanish, transl. Sampling Science). Bubok. Madrid.

Exciting replies interesting, I agree with most of the answers and explanations above.

My book Ciencia del Muestreo will be published on May 2017.

http://www.bubok.es/libros/251372/CIENCIA-DEL-MUESTREO

If not, which is the reason of using Statistics to validate all medical research?

yes i think so

Many subjective statistics have been used in Medicine, as classic and Bayesian statistics, but there is an objective way to medical statistics: do not leave to the researcher opinion what is objective reality.

Yes, I believe that yes, but Statistics must not be destructive with the subjects to be respetable objective science.

Statistics and measures do not say all about objective knowledge in Medicine, but they cannot be omitted in the diagnosis and evolution of a patient or subject.

Yes, it can but this cannot be improvised.

Statistical knowledge can help to the doctor in medicine, but it is more important to love the patient and treate him as one would treate to oneself.

An article published about this area is the following one,

I recommend the fourth edition of my book "Fundamentos de la Inferencia Estadística Objetiva".

http://www.lulu.com/shop/mariano-ruiz-espejo/fundamentos-de-la-inferencia-estad%C3%ADstica-objetiva/paperback/product-23340232.html

Other reference is "Recientes frutos en bioestadística" published in Estadística Española 60(195) (2018), 61-84.