Bayesian voodoo and the COVID19 vaccine coverup
The Vaccine Adverse Event Reporting System (VAERS) has been in the spotlight recently, as the Centers for Disease Control (CDC) and Food and Drug Administration (FDA) squirm their way around attempts to get details about their COVID19 vaccine safety monitoring efforts. CDC Director Rochelle Walensky recently responded to such a request from a US senator with this letter, in which she said they used Bayesian analysis (which she calls Empirical Bayesian or EB data mining) to detect potential safety problems. What does that mean?
The Bayesian revival
Thomas Bayes
Bayesian analysis is a late-20th century revival of an 18th century probability theory that “fell into disrepute,” according to the International Society for Bayesian Analysis (ISBA). The reason for this fall from grace, according to the ISBA, was “because they did not yet know how to handle prior probabilities properly.” This problem was allegedly solved in recent years with “new computational methods” and “powerful computers,” resulting in newfound popularity for the once-disreputable Bayesian technique.
The ISBA is treading delicately around a fundamental problem with Bayesian analysis: it can easily be misused, whether inadvertently or by design. The user can make subjective choices that influence the results—it’s not necessarily based only on objective data, so the quality and integrity of the results depends on the quality and integrity of the user. This characteristic probably has something to do with not only the Bayesian fall from grace, but also with its newfound popularity in the Age of Fakery.
What is Bayesian analysis?
The CDC calculates the frequency probability of a death report to VAERS from a COVID19 vaccine (0.0027%).
Bayesian probability is in contrast to traditional frequency probability, which the CDC uses to calculate the probability of a VAERS death report, for example, and which I use on the CVax Risk page to estimate the risk of COVID19 vaccination from VAERS data. You could do the same calculations with any vaccine injury data you wanted to use. The results have their limitations, of course, but unlike Bayesian analyses, they are completely objective within those limitations, calculated from the number of vaccine injuries and the number of vaccines administered.
Author Roy Goldman wrote about the historical philosophical conflict between Bayesian and frequency probability:
“For hundreds of years, probability was thought to be a ‘doctrine of chances’ or ‘frequency.’ One repeats an activity many times (e.g., throwing dice) and then counts how many times a certain event occurs (e.g., throwing a 7 or 11 in a craps game). Some very famous statisticians that we all study […] were in this frequency camp. While they detested each other, their common enemy were the Bayesians, who believed that other relevant historical or prior information may be used to build model parameters. Sometimes a prior assumption was picked out the air, but it still led to useful outcomes. Such an assumption was considered anathema by the frequentists.”
Bayes’ Theorem. Image credit: towardsdatascience.com
Here is the ISBA’s clear-as-mud description of Bayesian analysis. They’ve lumped a lot into this long paragraph, but one crucial fact stands out: the results of Bayesian analyses are malleable, depending on the analyst’s input.
“There are many varieties of Bayesian analysis. The fullest version of the Bayesian paradigm casts statistical problems in the framework of decision making. It entails formulating subjective prior probabilities to express pre-existing information, careful modelling of the data structure, checking and allowing for uncertainty in model assumptions, formulating a set of possible decisions and a utility function to express how the value of each alternative decision is affected by the unknown model parameters. But each of these components can be omitted. Many users of Bayesian methods do not employ genuine prior information, either because it is insubstantial or because they are uncomfortable with subjectivity. The decision-theoretic framework is also widely omitted, with many feeling that statistical inference should not really be formulated as a decision. So there are varieties of Bayesian analysis and varieties of Bayesian analysts. But the common strand that underlies this variation is the basic principle of using Bayes’ theorem and expressing uncertainty about unknown parameters probabilistically.”
What have the CDC and FDA been doing?
Starting in April 2021, I questioned how it was statistically possible that there had not been any alerts from the unprecedented number of VAERS reports, given the data was supposedly being monitored for “disproportional” reports, and reports received “at least twice as frequently as expected.”
ACIP presentation Jan 27, 2021, Page 15.
At that time, the most recent COVID19 vaccine safety monitoring information was a January 27 presentation from the Advisory Committee on Immunization Practices (ACIP), which looked at VAERS data for approximately the first month of the rollout.
On Page 15 of the ACIP presentation, we’re told that VAERS monitoring using Bayesian analysis was designed to identify adverse events reported “at least twice as frequently as expected for a COVID-19 vaccine compared to the VAERS database.” We’re told these analyses were being run weekly and that no alerts had occurred as of January 22, 2021.
ACIP presentation Jan 27, 2021, Page 13.
Page 13 of the ACIP presentation tells us that 979 serious adverse events were reported to VAERS during the first month of the COVID19 vaccine rollout. The authors didn’t say how that number compared to historical VAERS averages. If they had, they would have had to explain why there were no alerts. Let’s look at the numbers.
Table 1: Serious events reported to VAERS. See Note 7 on the CVax Risk page for the source of this data.
Table 1 shows the number of serious events reported to VAERS from 2010 through 2020. The yearly number was divided by 12 to estimate the expected number of monthly reports. Average monthly reports ranged from a low of 148 to a high of 245, and the 10-year average was 182.
The number of serious injury reports from COVID19 vaccines in the first month, 979, was more than 5 times higher than average, but we’re told there were no alerts. How was that possible?
The number of serious injuries and deaths reported to VAERS from the COVID19 vaccines continued to escalate dramatically. By the end of 2021, VAERS was getting over 5000 serious adverse event reports from COVID19 vaccines each month—many times more than ever before in VAERS history.
The conspiracy of silence
While the CDC and FDA have acknowledged a few isolated risks of COVID19 vaccination such as blood clots and heart inflammation, they have yet to address—or even mention—the massive increase in deaths and serious injuries reported to VAERS from the COVID19 vaccines overall. The conspiracy of silence about that issue is still in full force among vaccine safety researchers who looked at VAERS data—none of whom have even acknowledged the unprecedented number of reports, much less analyzed the reasons for it.
Instead, we have research like this example from March 2022, in which the authors give us a “so what” conclusion in a stunning display of wasted time and money—because who cares if a lot of people are suffering devastating injuries and death, as long as most of them are not, right?
“Interpretation: Safety data from more than 298 million doses of mRNA COVID-19 vaccine administered in the first 6 months of the US vaccination programme show that most reported adverse events were mild and short in duration.
“Funding: US Centers for Disease Control and Prevention.”
Possible explanations
There are a few possibilities I can think of as to why there were no safety alerts in VAERS from Bayesian data mining right from the beginning of the COVID19 vaccine campaign:
The Bayesian analyses were programmed to expect an extraordinarily high number of injury reports from COVID19 vaccines, resulting in no alerts.
The analyses excluded a large number of VAERS reports, for unknown reasons.
The analyses weren’t done at all.
The analyses did result in alerts, but we were told otherwise.
In any case, the answer can be found by examining the FDA’s Bayesian analyses—which the FDA has apparently refused to provide, in response to a recent FOIA request. Go figure.