our covid model says "X saves lives"
passing off a model as evidence is tantamount to lying
it’s honestly both jarring and disheartening that even at this stage of the game the same shabby tricks keep getting trotted out to try to make the abject failures of covid vaccines appear to be victories.
they have slanted data, rigged studies, used mice to stand for people and biomarkers to stand for clinical efficacy. truly, the whole panoply of prevarication has been deployed to try to cover for the basic fact that these products barely worked when they were launched, likely only ever had positive risk reward in extremely high risk people, and given their leaky/non-sterilizing nature were always going to rapidly drive viral escape and viral advantaging as hoskins effect/OAS set in.
this not only inverted efficacy but likely has large negative societal effects overall. i doubt you could make a case for these “vaccines” in even the highest of high risk categories anymore. they’re pretty much all negative net value on risk/reward for pretty much everyone.
and this would seem to be why they are no longer even trying. why seek data when you can simply conjure efficacy by going back to that most egregious of fabulism follies: completely making stuff up?
come now señor cat, surely this is too harsh an indictment!
well, let us just see about that, shall we?
this new “STUDY” has been getting a lot of press.
it certainly seems an impressive set of claims to wave around.
but here’s the thing: it’s not a study at all.
it’s just a model pretending to be data.
it’s literally just self-referential circular assumption bias being passed off as fact by making it sound “sciency”
Rather than modeling population-level effects, the team of researchers from the University of Maryland, York University, and the Yale School of Public Health used a computational model that allowed them to incorporate factors like waning immunity or different age groups’ eligibility for vaccines and boosters into their calculations. Fung noted that adding all of these parameters together creates more statistical uncertainty in the data, meaning that there’s a larger margin for error in the study’s final results. The authors acknowledge this uncertainty by providing “credible intervals” for their calculations — ranges that show, for example, that their estimates for the number of averted deaths would be between 3.1 million and 3.4 million.
what they are basically saying here is that if you make a whole bunch of small assumptions and then multiply them by one another, you get BIG error. that, at least, is honest.
this however, is not:
To calibrate their model, the researchers first made sure it could correctly predict actual case, hospitalization, and death patterns for the December 2020-November 2022 time frame. They then removed the vaccination elements of the model to examine what would have happened without the Covid-19 vaccines.
this is the oldest trap in hindcasting. if you took this claim to any wall st desk or HFT shop, they would be on the floor laughing. it is hilariously easy to build a model full of plug variables or odd inflections/assumptions that fits past data. you just lock in the presumed effect of a couple things you care about and then bend everything else into shape around it.
it is so rare that such a model would then go on to exhibit future predictive value that pretty much no one even bothers trying this sort of approach anymore unless they are either woefully ignorant and about to learn about how expensive lessons in finance can be or they are trying to fleece the rubes.
neither is a moneymaker. (at least not for the rubes)
epidemiology is no different.
excursions into nonsense
“The model incorporates the age-stratified demographics, risk factors, and immunological dynamics of infection and vaccination. We simulated this model to compare the observed pandemic trajectory to a counterfactual scenario without a vaccination program. "
adding vast complexity to a nonsense methodology does not even out into some sort of reasonable average. it just gets you complex nonsense with far greater potential for radical error excursions.
this looks very technically impressive
until you remember all the other models just like this that got trotted out over and over in covid and backfit perfectly and then fell flat on their faces the minute they tried to make forward predictions because (repeat after me): ability to hindcast says NOTHING about ability to forecast.
this was the same sort of “model” and “variance measure from assumed mitigation.”
the green line (not the black one) was what actually happened.
i could pull these all day, but i doubt there’s much point. from SAGE to the UW, we saw this same massive over-estimation trend again and again.
and it’s now being used to validate vaccine effectiveness.
but this is a wholly invalid means to assess such a thing. all you’re really seeing are the assumptions baked into a complex model that we’re not allowed to see or assess or run ourselves.
and it was full of bad assumptions.
they did not even calculate vaccine efficacy or look for curve bending or symmetry violation. they assumed it from exogenous research.
and that was the whole game right there.
we all saw the shenanigans with ignoring immune suppression windows, denominators rigging, population slanting, and 20 other games played on “VE.”
Vaccine efficacies against infection, and symptomatic and severe disease for different vaccine types — for each variant and by time since vaccination — were drawn from published estimates.
just by mathematical necessity, they are clearly assuming VE’s in the 85%+ range. it’s the only way to get to this kind of number given the 70% US vaxx rate (or even 90%+ in high risk) over such a short interval and when around half of US deaths occurred before vaxxes hit even 20% penetration.
(and, obviously, no vaccine works before it is taken)
think about what this means.
let’s assume they are assuming 85% vaccine efficacy (they likely went higher, but this is for illustration). it’s a built in parameter. so if you take it out, deaths spike by 6.7X. this is programmed into the model. it’s not some law of nature. it’s just an assumption based n studies that have long since been invalidated.
the whole rest of the backfit takes this into account. all the assumptions about viral attenuation, population dynamics, resistance from prior infection, you name it. and if your 85% is too high, in there somewhere, you either missed a huge variable or you vastly underestimated one. if you didn’t, it wouldn’t backfit.
so, if it does backfit and your major parameter assumption is wrong, it means the whole rest of your model is garbage.
you have loaded this model to show you what you presumed when you run the “counter factual no vaccine case.” it’s pure GIGO and the minute you assumed “vaccines worked well” then “vaccines saved huge numbers of lives” will pop out.
but if this assumption is wrong (as it appears so clearly to have been in the israeli palestine natural experiment comparison where death rates in the two places were near indistinguishable both before and after vaccination despite wide divergence in vaxx rate) then you’ve just “proven” nothing at all apart from the fact that models express the assumptions of the modeler.
wanna bet this model cannot backfit both israel and palestine while using the same parameter assumptions?
because that would be an interesting test.
these guys literally read the pfizer marketing materials and presumed them true (despite the fact that they clearly never were). even 50% functional risk weighted efficacy early on was a stretch and that eroded rapidly as virus evolved to be vaxx advantaged and to infect the vaxxed at far higher rates.
by the time we hit omicron, we were showing 3-4X the infection rate in the vaxxed vs unvaxxed in the UK data which was the best set i ever saw globally.
and that’s a HUGE effect as a parameter input.
it means that 50% VE in any given case is -50% to -100% functional VE at society scale once you add in the greater risk of infection. (if you are 50% less likely to die if infected but 4X more likely to get infected, your overall chance of death doubles etc)
it’s not hard to see why they stopped reporting this back in may…
(data in graph taken directly from UK.gov data. methods and links to sources HERE)
i would be willing to wager a whole bag of cat treats that this was not included in the “model.”
quite a potential effect to leave out.
this is the other problem with trying to model systems this complex. you have no idea what you left out and until you can successfully make forward predictions, you have no idea if you added enough to actually be skillful in prediction/assumption. you’re just chasing your tail in circular logic games.
i notice they also failed to include any costs from vaccine injury or death in the calculus on lives or expense. (odd for drugs that showed no all cause mortality benefit in their own trials and with so many known and severe adverse events)
and this “model” they are using is WAY out of step with the actual data from places like NY (which has been so highly representative and looks like the whole northeast)
hospitalization in 70+ age group is up about 50% from a year ago and has been far higher all summer. this group is over 95% vaxxed in NY.
this curve is bending in the wrong direction.
from this data, one might well ask: are the vaccines making it worse?
it’s far too confounded a series to prove such a claim, but it would be more consistent with the pattern than “vaccines are working” especially given the much milder variant and that we certainly saw less of this effect in lower vaxxed states.
it’s time these “reality diverged from our model therefore the thing we programmed our model to think works must have worked” people were put out to pasture.
this is cynical soundbite sensationalism made for media not meaning.
this is not science, it’s science fiction as propaganda paid for by the uber-ESG partisans at “the commonwealth fund” who have been pushing such dodgy science all covid.
if this is the best data they can muster, then they have just admitted defeat.