Or higher numbers if they are being paid by the government for every Covid case the hospital treats.
3 Likes
That would never happen, Iām sure.
Well, they didnāt claim George Floyd as a Covid death, so thereās that.
3 Likes
RFK Jr. is an inspiration to all āBoomersā that lift !
1 Like
Iāve been saying for a long time that people donāt want freedom. They just want to be irresponsible.
Iām pretty sure I know more about statistics than you do. But thanks for making personal comments based on a complete lack of knowledge. No doubt this is what the greatest statisticians do - commit type II errors by failing to reject their null hypotheses. I donāt want to be mean, but you might not even be an average statistician? (We could do a Ļ2 test on your independence and goodness of fit).
You are right different countries provably have different methods of counting deaths or diagnosis. But ratio of deaths to expected deaths overcomes some of these calibration concerns. Which are hardly insurmountable obstacles - meta-analyses look at studies with different parameters and can still draw mathematically sound conclusions in many cases with higher quality studies. It is true the errors get higher and the significance lower. It is not true no meaningful conclusions could be drawn about Covid.
Convince me.
Three statisticians go hunting. The first one fires his gun at a buck, but the shot is off to the left by three feet. The second statistician fires his gun, but his shot is three feet to the right of the buck. The third statistician gets excited, and yells āGot it!ā.
Not going to try to convince you - odds are Iām just some guy on the Internet. But I remember having to do this famous problem (Buffonās needle, from 1733) as one of our exam questions for stats class, as I am sure you probably did as well.
If the data analysis yields a P-Value of greater than 0.05, I like to say that there isnāt enough data to reject the null hypothesis. Sure there could be Type II Error, but we just donāt have enough data to accept the alternate hypothesis.
My training and experience in statistics is in Six Sigma and assisting others achieve a Green Belt. I spent my strongest focus on the Measure Phase of the DMAIC process.
In College I took no statistics but more mathematics than anyone I know. Being involved in the Six Sigma initiative at my work was centered on real world application of statistics.
I donāt know that much about business standardization protocols. But I took over a dozen university level math courses, one at the masters level. One in statistics, one in numerical methods and analysis. I TAād three math courses at a decent uni, including complex variable calculus.
Couldnāt agree more
Probably wouldāve worked better if people just had faith it would. I mean, believed The Science.
No serious person believes that Covid was handled either well(unless theyāre talking about relative to countries that handled it like a big, society-wide threat) or shouldāve been handled more aggressively.
We might have a similar mathematics education.
On top of the three Calculus courses that were required for Nuclear Engineering, I took two Advanced Calculus courses plus a Mathematical Analysis course to help raise my GPA. Nuclear Engineering required not only elementary differential equations, but also a course in intermediate differential equations (those equations with variable coefficients).
I taught myself statistics via Minitab and books, though I did have a statistics course in high school when all we had was a slide rule for calculating standard deviations (1965). For that reason I avoided all statistics courses in college. Talk about tedious work. No thanks
I firmly stand that when comparing apples with apples they must be measured by the same method and criteria.
This is at the foundation of Six Sigma, which is real world application of statistics
In the real world you gotta work with what you got. Sure, if everything was standard in every country it would be much easier. But you can compensate for most of the obstacles and differences, provided the numbers are not given in bad faith. Clearly sometimes the numbers are so far from average that they have almost certainly been cooked. You can essentially prove that when the numbers are way off, as with some countries, and exclude them. But that doesnāt give you the real story in those specific places.
Many newer medical studies use āmeta-analysesā where they take results from perhaps dozens of other studies - pooling sample sizes to increase statistical power. These studies have the same focus (say, do dietary salt levels affect blood pressure?), but different sizes, age groups, comorbid conditions, medical confounders, measured variables, quality of research, techniques, instruments, methods, types of error, formats (case-control versus blinded versus retrospective analysis, etc.). You can still mathematically compensate for much of this, and make meaningful conclusions. But not if most of the studies just make bullshyte up. Garbage in, garbage out.
If you had followed me helping no more than three people with the Measure Phase of their Six Sigma project, you would surprised the amount of variation in the Gage in the Measurement System Analysis. In a Gage R&R analysis we are looking for high part-to-part variation relative to the variation in the Gage. (Gage is a Six Sigma term used for what measures the Y that the project is attempting to improve.)
With Covid deaths there was no autopsy plan in the USA to determine cause of death. That in itself places loads of variation in the Gage. Was dying in the hospital and also testing positive for Covid all that is needed for the death to be classified as a COVID death? Might this have varied hospital to hospital. Even trying to do the right thing, the Gage variation is high.
Add to that, the government paying subsidies for every Covid case treated in the hospital. It is a recipe for a terrible Gage.
Sure, Covid might have caused lung failure or whatever, and may or may not have been formally tested for, especially early in the pandemic, More so if the death was not in hospital. Payments and politics skew results.
But you can still get meaningful information just by comparing the number of recorded deaths in a given place to the number expected there at a certain time based on previous experience. These records usually go back many years, and there is more than enough data to work with. Even wastewater analysis yields some useful information.
You wonāt find a much bigger fan of the hard science and medical research establishment than this layperson, but those guys sure blew it when their number got called to step into the public policy limelight. Not through any fault with the underlying methodology, but through the hubristic attempts to control peopleās behavior with bad public policy that backfired and squandered credibility.
It would seem to me that thereās an awful lot of noise in an awful lot of COVID data. Politician, activist and grifter fingerprints are everywhere, but Iām a mere six sigma green belt who demonstrated a good grasp of calculus 25 years ago.
That stuffās all gone now, off to the same nether region of my brain that all of those phone numbers I used to know now reside.
If thereās anyone with a chance of bringing us simpletons back into the fold, its an RFK type holding a few folks accountable. The people who might be upset by an RFK type in this role arenāt the ones who might cause future problems through an irrational distrust of scientists.
Unfortunately for scientists in 2024, there are a lot of rational reasons to not trust them that have nothing to do with the method.
4 Likes
Iād be wary of attempting to draw any firm conclusions with respect to Swedenās mitigation efforts (or lack thereof). There appear to be far too many confounding variables (e.g., worldwide leader in single-person households per capita, high vaccine compliance, natural social distancers, socialized healthcare, etc.). Furthermore, while they definitely did far less than most other countries, there were still restrictions placed on the number of attendants/participants in public gatherings (dropped to no more than 8 in 11/2020), no private visits to nursing homes, secondary schools switched to distance learning, and table-service only at restaurants.
For what itās worth, Iāve also come across statistical analyses from quants/actuaries attempting to show that, when one takes into account Swedenās downward trajectory in mortality over the span preceding the pandemic, they actually fared poorly in comparison to many other countries, including their Nordic neighbors. To be fair, though, Iāve certainly seen arguments to the contrary and will readily admit that I donāt have the statistical acumen to evaluate either credibly.
Side note: I distinctly recall being rather amused at the complete flip-flopping that occurred regarding Sweden. For years, those on the left would point to Sweden (and other Nordic countries) as evidence for supporting single-payer healthcare, prison reform, and a more robust social safety net. Conservative folks generally responded by pointing out the racial homogeneity and smaller population of Sweden. Then, Covid takes center stage and the sides walk right past each other on their way to the opposite ends of the spectrum. People on the right want to emulate Swedish policies while the left repeatedly emphasizes the confounders.