albert einstein graphic

Probability Science - in medicine and in physics

Probability Science graphic HOME .... William Gilbert . Rene Descartes . Isaac Newton . Albert Einstein ........ String Theory ........ General Image Theory Probability Science graphic
probability science dice photo

Probability is increasingly used in many areas of modern science, and most notably in medicine and in physics, in scientific proof claims. But often probability is used poorly in science and really gives little or no proof of what is claimed.

In medicine, probability is now commonly used in survey data analysis, as where a 10% correlation between peoples illness and peoples behaviour statements is said to prove eg that general behaviour A always has a 10% risk of causing illness B. Often the general behaviour A has no actual effect on the illness B, but has some correlation with the use of some unidentified product C which is the actual cause correlating 100% with illness B. But the incorrect medical claim is pushed.

In physics, probability is now commonly used in experiment data analysis, as where a 95% correlation between photon emissions and some general magnetic event is said to prove eg that general magnetic event A always causes photon emission B. But the general event A may have no actual effect on the emission B, but has some strong correlation with some unidentified specific event C which is the actual cause correlating 100% with emission B. Probability is also used as the basis of Quantum Mechanics.




Probability science graphic

In science today probability is widely used in different aspects of data analysis proof claims that are not reviewed by statisticians. It is used in experiment data analysis and in survey data analysis, and in both areas it is also used in error estimation. But probability is commonly used wrongly in science as noted by some major statisticians like R.A.Fisher. It is commonly used by amateur-statistician 'scientists' who are not good statisticians and consult no statistician, so much that the journal 'Nature' has now started asking statisticians to review some submitted papers.

Probability in Medicine

In medicine, probability is now used both in experiment data analysis and in survey data analysis but here we will consider chiefly the latter (below under Physics we will consider the former). The chief problem with survey data is that it always involves some limited number of selected people being asked some limited number of selected questions. It may be that an illness being studied is caused by ACME soap, but the survey had no question about ACME soap or it did but none of the people surveyed used ACME soap. But still that survey will be probability-tested for that illness, and may well give some correlations for that illness. It will be announced that some behaviours 'are a risk for the illness', while ACME soap passes unmentioned.
(PS. this is NOT a claim that ACME soap causes any illness, we use the name here only as the name of 'some hypothetical product'.)

We can now consider a hypothetical medical survey to be probability-tested regarding a hypothetical disease A ;


A hypothetical survey probability testing.

Where the unknown facts that the study seeks to discover are,
Disease A is actually caused by using too much of product C or the weaker product D.
Product C is more expensive than product D.
Product C is used more by middle-class vegetarians.
Product D is used more by working-class smokers.
Product C sells in less locations than product D, and some locations sell neither.

And where,
The survey is of pedestrians half from location X and half location Y, questions being ;
Do you own a TV ?
Do you regularly smoke cigarettes ?
Do you regularly smoke cigars ?
Do you usually drink more than 4 units of alcohol a day ?
Do you usually eat more than 2 eggs a day ?
Do you usually visit a gym more than once a week ?

This survey actually asks nothing about product C or product D, but will still give correlations for the illness caused only by these products as long as some people surveyed use either product. Hence,
TV-owners, cigar-smokers and gym-users on average may have higher incomes and to differing extents may buy more of the expensive product C than product D.
Egg-eaters on average may have the highest use of products C and D, and may have lower incomes and so buy more of the less expensive product D.
Alcohol-drinkers on average may tend to buy neither product C nor product D.
Location X on average may be more middle-class and sell more product C.

Survey question answers are used to split the survey population into sub-populations as 'TVowners' and 'non-TVowners'. Then probability testing may be done on illness rates between SOME answer sub-populations, when it should be done between ALL of the answer sub-populations - eg. between cigarette-smokers v non-smokers AND between cigarette-smokers v cigar-smokers AND between cigar-smokers v non-smokers etcetera. Of course a survey with many questions can give thousands of sub-populations, and while all should be probability tested, it is proper enough to publish the results only for all cases that exceed some specified significance level. (Alternatively probability testing can be done on illness rates between each answer sub-population and the total survey population, though that will dilute the probability differences and so can hide significant results)

Illness rates will vary between sub-populations, such that it may be reported that 'cigar-smoking carries a 20% risk for this illness' and 'egg-eating carries a 15% risk for this illness'. Of course in this case we know that these behaviours do not at all cause the illness - products C and D cause the illness. So the 'scientific truths' that this study claims are not actually truthful. Whence the saying that 'There are lies, damn lies and statistics'.

That holds even when probability studies are done properly, but often they are not. Hence cigarette-smokers v cigar-smokers may give a non-significant 5% while cigarette-smokers v non-smokers may give a significant 11%, but the study may have omitted to get or to publish the latter result. (as was the case for even the acclaimed Doll and Hill 1956 smoking survey study in regard to reported cigarette-lighter use -
Doll R, Hill AB (1956) Lung cancer and other causes of death in relation to smoking. Br Med J 2: 1071).

If there is some strong evidence for any hypothesis, then additional weak evidence will now commonly be taken as confirming and strengthening that. And even if there is only weak evidence for an hypothesis, then additional weak evidence will now commonly be taken as confirming and strengthening that. But logically only strong evidence should count towards proof, and weak evidence should only ever count as an indicator of a need to look for strong evidence. Generally there are no 20% causes and so no 20% risks, mostly A actually cause B or actually does not cause B. There may commonly be dose effects, and more rarely there may be multiple causes. But much too commonly medicine is reporting, and governments spread concerns about, relatively low illness 'risks' that are not actual scientific truths like 'eating fat causes heart problems' - and scientific journal 'peer review' has tended to create and keep backing such false discipline-prejudices. They might do better having chemists review physics papers, physicists review chemistry papers and astronomers review biological papers because their discipline-peer-review just promotes prejudice science instead of real science.

Medical research in the last 50 years has often centered on the use of statistics as shown clearly in nutrition studies. Hence a couple of published nutrition studies claimed to prove that Antioxidants were very good for peoples health, but then two later studies claimed to prove they were good for younger peoples health but were bad for older peoples health. And statistics-based claims were pushed strongly for a long time that margarine was better for health than butter, so lots of people have taken to using margarine. But new statictical studies now claim that margarine is worse for health. This bad science is likely killing people, yet todays 'scientists' and governments push it regardless. Many other published statististics-based nutrition studies have made doubtful claims due to poor use of statistics and there has been little if any really good nutrition research in recent years.

Probability in Experimental Physics

Probability testing in Physics and Astronomy is now commonly used in experiment data analysis or observation data analysis. This can have some of the problems seen in the use of probability testing of survey data. Hence where surveys can have omitted questions, experiment or observation can involve omissions in the factors investigated and this may have great impact in the more contentious areas like Particle Physics and Astronomy as with partial correlations between A and B being claimed as a causal proof when the true cause C was never studied.

Probability is also widely used in accuracy estimation, but often ignoring the probability fact that of several experiments or observations it is often NOT the one with the best accuracy that gives the most reliable evidence. Other significant issues are often also involved.

More recent Physics and Astronomy theories also commonly try to incorporate aspects of probability theory, correctly or incorrectly. Deductive assumptions involving infinities or limits often give false answers. So theory handling the infinitely small and infinitely large can ultimately require that the sum of an infinite set of zero probabilities add to a probability of one, which is plainly false. Physics deductions about the infinitely small or the infinitely large can generally be valid only derived correctly relative to some well proven specified finites. More recent physics theories can often involve error related to this issue.

False probability deductions can be due to a failure in specifying the data involved, or to a failure in specifying the assumed prior information involved. So there often can be no valid probability comparison between two physics theories regarding given data, if both involve assumptions about eg 'mass' but both fail to specify the prior information properties of 'mass' that their theories involve. Or phenomena that seem probabilistic may simply have some unseen or uncomputed non-probabilistic causes that may be currently unseeable or uncomputable.

Statistics based 'experiments' commonly rely on computer analyses or computer 'models' that are not fully specified and so such 'experiments' are not fully replicable to verify them or to challenge them. And replicable experiments generally though involving one set of statistical probabilities are all capable of being interpreted differently in terms of different theory paradigms. But statistics often cannot offer any valid evidence as to correctness between several alternative interpretations of an experiment. When radioactivity was discovered it soon became described as a 'causeless', 'random' or 'probabilistic' physical phenomenum, as no immediate cause could then be identified for radioactive events. But radioactivity includes nuclear fission which occurs naturally on Earth in Uranium and Thorium ores as 'spontaneous fission' and which was found to be caused by neutrons, so other radioactivity may well be caused by eg simultaneous neutrinos or other as yet unidentified causation and not really be 'probabilistic'.

Probability in Physics Theory

For some physicists the two-slit light experiment was taken as supporting a Heisenberg probabilistic quantum mechanics, as where there is some probability that an object actually at a specified time occupies one space location and actually at the same specified time in contradiction occupies some other space location. In such a probabilistic physics universe, the universe actually behaves probabilistically whereas in a determinate physics the universe actually involves fully specifiable causes giving fully determinate effects though that may not always appear to be the case. Probabilistic physics claims to be also backed by other supporting evidence, with claims of microscopic quantum processes such as 'superposition', 'entanglement' and 'virtual particle exchange' being involved.

Heisenberg's Uncertainty Principle basically assumes that all possible ways of determining an objects motion and position at some instant must involve changing the objects motion or position. But the Rudolphine Tables of Kepler allow determining the position and motion of a planet at some instant by calculation alone (which has no impact on the planets position or motion), and the position and motion of a body continuously emitting light can be determined for some instant from its emitted light signals (having no impact on body position or motion but maybe limited by light having a quantal nature). It seems that there will be some cases where such determinations in principle cannot be done accurately, but also that there will be some cases where such determinations in principle can be done accurately.

Some physicists do not support probabilistic physics including Einstein who rejected probability physics "because God does not play dice" (though that is maybe no scientific disproof and Einstein still accepted duality contradiction physics). Probabilistic physics is rejected also by others like Schrodinger who reject all contradiction physics, including Einstein dualism, as in his Schrodinger's Cat probability-exposing 'thought experiment' which is perversely often quoted to help 'explain' probabilistic quantum physics. But for those who reject contradiction in science, it exposes probability physics as contradiction nonsense. Yet for those who accept contradiction in science, it helps explain probability physics !? Of course it can be said that any claimed evidence for a contradiction must be contradictory evidence, and contradictory evidence may reasonably be taken as not being valid factual evidence - eg evidence that Jane is in Paris now AND that Jane is in Tokyo now or evidence that Jane is alive now AND that Jane is dead now ?! Logically it would seem that 'evidence' for a contradiction must be data being misinterpreted. It may be more scientific to say that nature itself is NOT probabilistic, but that human consideration of nature IS probabilistic and so can make nature APPEAR to be probabilistic. But nature showing apparent statistical associations will often allow of multiple alternative causal explanations or Image Theories, and in some cases necessarily do. See http://psych-networks.com/theoretically-distinct-mechanisms-can-generate-identical-observations/?utm_content=buffercae71&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer and http://www.new-science-theory.com/general-image-theory-1.php And of course A having a 10% chance of causing B, is also A having a generally or often ignored 90% chance of not causing B !

Probability methods generally are widely used in particle and quantum physics and have some use in almost all areas of physics today, even by physicists who reject actual probability physics. But where it is claimed that it has been proved that some physics is probabilistic, it is maybe best taken as meaning that it has really at most been proved that it is either probabilistic OR involves some as yet unidentified non-probabilistic causation. 2014 sees Christopher Ferrie and Joshua Combes, supported by Rainer Kaltenbaek and Franco Nori, throwing major doubt on Quantum Mechanics and especially its 'weak measurement' as being based on bad statistics. (see http://physicsworld.com/cws/article/news/2014/oct/09/are-weak-values-quantum-after-all)

While arguing for one-theory-only science, E.T.Jaynes concluded that probability theory has 'been fooled by a subtle mathematical correspondence between stochastic and dynamical phenomena'. But that rather supports multiple-theory science like Newton blackbox-theory science or perhaps preferably our General Image Theory science. See http://bayes.wustl.edu/etj/articles/prob.in.qm.pdf

Some of these physics probability issues were considered at the CERN 2007 conference 'Statistical Issues for LHC Physics', see http://physicsworld.com/cws/article/indepth/43309 Many suggest replacing the long-standing use of a probability value (p-value) of below 0.05 for 'significant' results with a stiffer p-value threshold of maybe 0.005, which should help to improve the use of probability in some areas of science though this does not affect the other issues with probability science. The probability of the Sun tomorrow not rising in the East and setting in the West is below 0.00000000001 but even that does not prove that the Sun orbits Earth daily, as used to be commonly believed though now we know Earth revolves daily. Probabilities are probably often best used just to help identify specific issues where further real experiment are more likely to be useful. Of course misuse of statistics is far from the only problem with science but hard-science Physics is the leading edge of science, unfortunately long leading in bad science only worsened by bad use of probability mathematics.

Mathematics is helpful to science chiefly insofar as it can help to increase exactitude in both experiment and reasoning proofs, but probability mathematics is basicly the mathematics of inexactitudes and so really can only help show the extent to which science proofs may be uncertain. Probabilities cannot themselves be causes of anything nor alone be proofs of any causations. And, without accepting Einstein's physics, the preponderance of science evidence supports laws of nature concerning nature not being probabilistic or playing dice - despite some apparent evidence for some seemingly contrary phenomena. Information is now commonly wrongly defined in relation to uncertaimties or probabilities but signal science shows many cases of information signals causing effects, and not lack of information uncertainties or probalities, as was perhaps well demonstrated in William Gilbert's 1600 'De Magnete' or 'On The Magnet'.

Probability science graphic


You are welcome to link to any page on this site, eg http://www.new-science-theory.com/probability-science.php


If you have any view or suggestion on the content of this site, please contact :- New Science Theory
Vincent Wilmot 166 Freeman Street Grimsby Lincolnshire DN32 7AT UK.


OR maybe make a small donation ;
(it will help with site development, and just possibly with some experiments long planned but never afforded.)

© new-science-theory.com, 2017 - taking care with your privacy, see Sitemap.