Plausible Reasoning 13
Lucky #13
In this final installment, I’ll do some summary, provide one, okay, two more examples of frequentism gone wrong, with one example I find instructive from the Covid-19 war crime ahem, “pandemic.” I’ll attempt to explain why “The Science” is NOT simply an accretion of “knowledge” and why the “Laws” of Physics don’t actually rule anything. (Quotes and italics here are used to make a particular point - as in the difference between fresh fish and “fresh” fish.)1 But I’m going to start by sharing a hypothesis that I saw presented at a recent lecture on this subject, about a phenomenon that I described in PR 4, that I think may shed some light on why we continue to have… the (ongoing) replication crisis, iatrogenic deaths in the “healthcare system” disease economy amounting to something like 400,000 people per annum killed in America, and massive pharmacological and broader scientific fraud, all produced by “The $cience.”
I. Certainty versus Uncertainty - a hypothesis.
At an early orientation lecture in law school back in 1996 - one of those ones you get in the first few days, where families are invited to come along and listen - either the Dean or one of the senior 1L professors gave a brief lecture. He told us that there had been a (then)-recent psychological study by some prestigious law school or institution, maybe the testing body for the LSAT or the ABA, and the upshot of it was that the single trait that determined whatever the metric was [success in law school, passing the Bar, success in law practice, satisfaction with career, etc.] was not LSAT score, or undergrad GPA, or anything quantifiable like that; rather, it was the ability to tolerate ambiguity. At the time I remember thinking that was an awfully odd claim - not at all what I had expected. What place would a comfort with uncertainty have in the law, I wondered? Were people not either “guilty” or “not guilty” before the law - a nice binary distinction - or liable or not liable for a civil wrong?
I hope that if I have conveyed nothing else throughout this discourse that I have imparted the necessity and significance of plausible reasoning to the law and to science, with a great nod to Francis Bacon’s initial work in defending and grounding both in induction. Subsequent developments in that same philosophical and intellectual tree - including by Henri LaPlace, Harold Jeffries, George Polya, Claude Shannon, R.T. Cox, David Stove, and Edwin T. Jaynes,2 and many other, less known and named intellectual heirs - have yielded significant advances in Information Theory (Shannon) and Probability Theory (Jaynes): they amount to a quantitative method for dealing with our ignorance and uncertainty in scientific discovery.
In one of the early installments in this series, I mentioned a “schism” in science:
…reminiscent of the one in the Catholic church, but this one begins with David Hume’s inductive skepticism, owes some of its intellectual origins to European upper society’s fascination with “games of chance,” but it truly manifests in the American body scientific as a result of the work of two academics: Karl Popper and Ronald Fisher.
I’ve gone over and over various manifestations of frequentist science in law and society more broadly, but what I didn’t cover fully was the why of this schism.3 I started with Hume’s inductive skepticism, but that in and of itself hardly seems sufficient to have driven this wedge into the House of Science, causing it to break into two rival gangs fighting for control of what Science is.
I was at a lecture this past weekend in which the central thesis was that this dichotomy is the byproduct of what seems to be a psychological predilection for certainty that avoids the emotional and intellectual discomfort that comes from uncertainty. You can sort a number of issues that I’ve discussed in these preceding 12 installments into seemingly dichotomous rival camps that look something like this:
Psycho-social Certainty Uncertainty Logic & Knowledge Deduction - Hume Induction - LaPlace Science Demarcation Falsification - Popper Predictive Strength Probability Theory Pr (D|H) - Fisher Pr(H|D) - Jeffries Statistics Frequentist - Venn Bayesian - Shannon Physics & Universe Ontological - Bohr Epistemic - Jaynes
Looking down the left-most column, what you see are all of the ways in which one’s dislike of uncertainty will likely draw one toward particular views of Science that appear to provide more certainty. The “data” obsession and the infinite dice-throwing [non]-“empiricism” of Fisher won out over the prediction of experimental results of Laplace, all of which led inevitably to the conclusion reified by vote at the 1927 Solvay Conference: that the Universe itself is indeterminate and can’t be known.
If you prefer the comfort of certainty, it’s very likely that you will find it hard to resist the siren song of that path, where the only knowledge that you can believe in is the closed loop of deductive knowledge. You will likely come to see science as being about falsifying prior theories, rather than understanding that new work can in fact limit or expand prior work, not simply divide it into a binary of yes (TOTALLY TRUE!) or no (TOTALLY WRONG!). Those with a penchant for certainty will find the probability of the data, given the anti-hypothesis (the null) to be more “objective” and disclaim that the probability of a hypothesis given the data even has meaning. As we took up in the more recent pieces, these two belief systems are with us today, in two co-existing but incompatible views of physics and the very structure of the universe, as being either indeterminate (i.e. probabilistic) and then “collapsing into existence” upon observation (the Copenhagen interpretation) - that the falling tree doesn’t make a sound in the woods if we aren’t there - OR that probability is instead epistemic, a rational measure of our ignorance and uncertainty about the universe… that God does not in fact play dice with Existence, and beavers chew down trees that fall in a noisome, crashing mess, whether we’re around to witness it or not.
II. A Parting Shot at Frequentist Science: Just How Risky Wasn’t Covid-19?
A. My Personal Hell Looks Like the DMV Office Run By Pharmaceutical Companies…
Near the absolute top of the most infuriating pieces of my professional and personal life are the vaccine mandates: not merely for all of the obvious reasons, nor that I’m against all forms of tyranny, great or petty, particularly the ones “sincerely exercised for the good of its victims.”4 In the case of Covid-19, it wasn’t just the government tyranny of health officials and their lickspittles, it was the cruelest of all tyrannies: technocrats claiming to follow “the science” while willfully ignoring their own rules, even when caught or splashed right in their faces. And I’m not even talking about the “getting caught unmasked” breaking of kayfabe that we all lived through. Worse even than that rank hypocrisy: imagine going to the DMV and having all of your paperwork in perfect order, spending weeks and months preparing, learning and re-learning arcane rules regarding motor vehicle transfers, taxes, and fees, and then when you’re called up to the window, upon proudly presenting all of your immaculately filled out paperwork, along with a money order in the exact amount necessary, the DMV employee looks you dead in the eye and tells you that none of those silly regulations mean anything. And then, at the moment your blood pressure starts passing through 130/90 on its way to never-before-seen heights, and you start sputtering in protest, she (unquestionably a she, either by biology or dress) taps her pen on a blue sign with white writing that says, succinctly:
“All Rules Subject to Change on Our Whim Because Covid-19. Thank you for your understanding and patience.”
THAT was my Covid-19 experience with the Food and Drug Administration (FDA). Yours too, whether you knew it or not. Let me show you what I mean.
B. …So Does Yours.
The following has always been - and still is - the FDA policy for health information providers:
Provide absolute risks, not just relative risks. Patients are unduly influenced when risk information is presented using a relative risk approach; this can result in suboptimal decisions. Thus, an absolute risk format should be used.
See Fischhoff B., Brewer N., Downs J. Communicating Risks and Benefits: An Evidence-Based User’s Guide. Food and Drug Administration (FDA), US Department of Health and Human Services; Silver Spring, MA, USA: 2011. [Google Scholar][Ref list].5
The National Institute of Health (NIH) funds and conducts biomedical research for the U.S. government, and is a very important sub-agency of HHS, alongside the FDA. They have a short paper explaining absolute and relative risk measures and how they are calculated. https://www.ncbi.nlm.nih.gov/books/NBK63647/
Relative Risk (RR) is simply the adverse outcomes (as a percentage of the test group) divided by the percentage of adverse outcomes of the control group. So, if a treatment had an adverse outcome of 15% in the test group and the same adverse outcome was 20% in the control group, the RR would be 15%/20% = 75%. This is to say that the higher the RR, the less effective the treatment happens to be. Should the RR exceed 100%, it means that the adverse outcome rates of the treatment surpass no treatment at all.
Relative Risk Reduction (RRR) is just 1, or 100%, minus the RR.
The problem with all of these rates is they omit the same critical information that is missing in the base rate fallacy - prevalence.
Absolute Risk Reduction (ARR) is meant to add this critical information into any intelligent risk calculation - or call it “knowing and voluntary” legal choice about your health if you like. ARR is the difference in the incidence of outcomes between the intervention group of a study and the control group. For example, and this is a HUGE mathematical problem for vaccine manufacturers, IF a large percentage of people in the unvaccinated group either don’t get the disease or get it and have no idea because it makes no impact on their life, then the ARR will not be great, even if the Relative Risk Reduction is high. More importantly, the Number Needed to Vaccinate (NNV), which is the inverse of ARR (i.e. 1/ARR) can turn the risk calculus ugly in a hurry, depending upon adverse outcomes from the treatment.
Let me put some concrete numbers to this:
The companies that made the various mRNA gene therapy products published data suggesting wildly successful “Relative Risk Reduction” (RRR) for their products. These numbers were 95.1% for Pfizer, and 94.1% for Moderna, and 67% for the Johnson & Johnson.
As was also noted in the BMJ Opinion, Pfizer/BioNTech and Moderna reported the relative risk reduction of their vaccines, but the manufacturers did not report a corresponding absolute risk reduction, which “appears to be less than 1%”. Absolute risk reduction (ARR) and relative risk reduction (RRR) are measures of treatment efficacy reported in randomized clinical trials. Because the ARR and RRR can be dramatically different in the same trial, it is necessary to include both measures when reporting efficacy outcomes to avoid outcome reporting bias. In the present article, a critical appraisal of publicly available clinical trial data verifies that absolute risk reduction percentages for Pfizer/BioNTech vaccine BNT162b2 and Moderna vaccine mRNA-1273 are, respectively, 0.7%; 95% CI, 0.59% to 0.83%; p = 0.000, and 1.1%; 95% CI, 0.97% to 1.32%; p = 0.000. The same publicly available data, without absolute risk reduction measures, were reviewed and approved by the roster of members serving on the U.S. Food and Drug Administration’s (FDA’s) Vaccines and Related Biological Products Advisory Committee (VRBPAC) for emergency use authorization (EUA) of the mRNA vaccines [10]. Ironically, the omission of absolute risk reduction measures in data reviewed by the VRBPAC overlooks FDA guidelines for communicating evidence-based risks and benefits to the public [11].
They intentionally omitted the ARRs for a reason - because it showed that the gene therapy products sucked - even as against the original strain for which they were designed (rather than the version extant by the time the products were approved, had manufacturing approved, manufactured, shipped, and delivered to clinics for injection into arms. Hence why all of that had to be vitiated by the EUA waiver of all inspection and other normal regulatory safety protocols.)
ARR is also used to derive an estimate of vaccine effectiveness, which is the number needed to vaccinate (NNV) to prevent one more case of COVID-19 as 1/ARR. NNVs bring a different perspective: 81 for the Moderna–NIH, 78 for the AstraZeneca– Oxford, 108 for the Gamaleya, 84 for the J&J, and 119 for the Pfizer–BioNTech vaccines. The explanation lies in the combination of vaccine efficacy and different background risks of COVID-19 across studies: 0·9% for the Pfizer–BioNTech, 1% for the Gamaleya, 1·4% for the Moderna–NIH, 1·8% for the J&J, and 1·9% for the AstraZeneca–Oxford vaccines.
ARR (and NNV) are sensitive to background risk— the higher the risk, the higher the effectiveness—as exemplified by the analyses of the J&J’s vaccine on centrally confirmed cases compared with all cases: both the numerator and denominator change, RRR does not change (66–67%), but the one-third increase in attack rates in the unvaccinated group (from 1·8% to 2·4%) translates in a one-fourth decrease in NNV (from 84 to 64).6
For every 1 person that received a benefit from the Pfizer BNT162b2, roughly 142 did not, depending upon data and confidence interval. The data for the Moderna experimental gene therapy is roughly the same, with roughly 1 in 100 people receiving a benefit from the shots (which is to say, 99 would not receive any benefit at all) - while in both cases, the people receiving the shot - the 99 or 141, would also at the same time be exposing themselves to some risk associated with receiving an unlicensed gene therapy product.
These experimental gene therapies also come with their own concomitant risk. For the adult recipients (age 16 and older), the Pfizer COVID-19 clinical trial found the overall incidence of severe adverse events during the two-month observation period to be 1.1%, or 1 in 91, which is larger than the ARR for the Pfizer experimental gene therapy. [FN76] When this phenomenon was further studied after the EUA was granted and injections were performed on the general public, it was found the rate of severe adverse events [FN77] went from 1:91 to 1:43, over double the trial rate.[FN78] This means that as a matter of relatively straightforward mathematics, the Pfizer “vaccine” is more than three times as likely to result in a harm to a recipient as it is to result in a benefit, which (as noted infra) requires 142 people to be vaccinated, before the 143rd person will obtain that benefit. According to the numbers, we should expect at least three people out of that same 143 to have had a serious adverse event by from being injected with the Pfizer shot.7
This is the other part that proves scienter by our technocrats: the omission of NNV’s could perhaps be chalked up to negligence, but the simultaneous willful suppression, manipulation, change in reporting criteria, etc., and willful ignorance regarding adverse events was designed to hide what Bayes’ Theorem’s use above was saying right from the beginning: the Covid-19 vaccines NEVER made any sense at all because they never passed their own risk-benefit analysis… you could reasonably expect three people to be harmed by some kind of adverse reaction before even one person would see any benefit. They simply ignored this, pretended it was all fine, even when their own numbers showed that it was all a lie. They were never “safe” and “effective” by any sane definition of those words.
A critical appraisal of phase III clinical trial data for the Pfizer/BioNTech vaccine BNT162b2 and Moderna vaccine mRNA-1273 shows that absolute risk reduction measures are very much lower than the reported relative risk reduction measures. Yet, the manufacturers failed to report absolute risk reduction measures in publicly released documents. As well, the U.S FDA Advisory Committee (VRBPAC) did not follow FDA published guidelines for communicating risks and benefits to the public, and the committee failed to report absolute risk reduction measures in authorizing the BNT162b2 and mRNA-1273 vaccines for emergency use. Such examples of outcome reporting bias mislead and distort the public’s interpretation of COVID-19 mRNA vaccine efficacy and violate the ethical and legal obligations of informed consent.8
A final note on just how insidious frequentists statistics can be in the law: in PR 9, regarding the Prosecutor’s Fallacy and Sally Clark’s wrongful conviction, using bad math to send a grieving mother to prison for a term of years, a reader of that article correctly pointed out that I had missed one of the greatest examples ever, and it even has its own wikipedia entry, but the California Supreme Court opinion does the job well enough to illustrate the problem, and it’s not a long or difficult read.
The conclusion of this whole effort could be summed up by observing that our entire experience is contingent and tentative: memento mori is a good place to start, but that we can leave behind no small or unimportant legacy through our own contributions to the objective branch of knowledge - science - by our understanding of what that means. Our models about how the Universe works are judged by their predictive power alone; maintain sufficient humility to appreciate that
…All that we have are models. The Universe does not have any concerns about such things as “distance” or “light years” or maps or mathematics or “strings” or “membranes” or “Copenhagen interpretations” or limit functions or even “the backs of turtles.” The Universe is far more complex than we can ever project onto it and it does not OBEY "our “laws” at all; rather, we make models that attempt to describe the Universe and the best ones have the best predictive strength. Notwithstanding all of that…
….We need not grope blindly in the darkness, but instead can use probability theory to precisely measure our ignorance and search for answers there, as Laplace advised and taught, but now we can do so using enhanced tools and understanding through E.T. Jaynes’s distillation of this inductive process into the “optimal processing of incomplete information.”
…which is another way of saying, it’s the fastest way to Learn, and you increase your chances of stumbling upon something that passes for genuine Wisdom - in either the Law or Science.
With a tip o’ the cap to David Stove.
And the inestimable (Dr.) Jeff Glassman.
Notwithstanding that I do have a Heading entitled “Why,” I really answered the “How” question, something I am wont to do by bent of personality. The how is very often the practical person’s interpretation of the answer to the why question. Why questions can be interpreted both ways: functionally through the psychological lens and also in the idea’s
“It would be better to live under robber barons than under omnipotent moral busybodies. The robber baron’s cruelty may sometimes sleep, his cupidity may at some point be satiated; but those who torment us for our own good will torment us without end for they do so with the approval of their own conscience. They may be more likely to go to Heaven yet at the same time likelier to make a Hell of earth. This very kindness stings with intolerable insult. To be “cured” against one’s will and cured of states which we may not regard as disease is to be put on a level of those who have not yet reached the age of reason or those who never will; to be classed with infants, imbeciles, and domestic animals.” ― C.S. Lewis, God in the Dock: Essays on Theology (Making of Modern Theology).
This is a variation on the same problem we tackled previously in Plausible Reasoning 8, regarding medical testing and the base rate fallacy.
Comment: COVID-19 vaccine efficacy and effectiveness—the elephant (not) in the room, www.thelancet.com/microbe Vol 2, July 2021. Originally published April 2021, then corrected, and republished online https://doi.org/10.1016/ S2666-5247(21)00069-0
Wilson v. Austin, 22-cv-438 ECF 1, Complt. ¶114, (May 25, 2022).
Brown RB. Outcome Reporting Bias in COVID-19 mRNA Vaccine Clinical Trials. Medicina (Kaunas). 2021 Feb 26;57(3):199. doi: 10.3390/medicina57030199. PMID: 33652582; PMCID: PMC7996517.