No, my analysis is correct as usual. If one treatment reduces deaths 99% and the other reduces them 97% then yes, one is 3-fold better. This is absolute true.
If your cardiologist is offering you two CVD medications and says: they both reduce your risk of a heart attack in the next year by >90% but patients on medicine A have 3x the rate of events than patients on medicine B, which one are you taking?
Small numbers matter!
Regardless, the death data by vaccination status between the two countries confirms my point.
Death rates between the two countries split out by unvaxxed/vaxxed/boosted are very similar. It's not like the US is seeing more mortality across the board due to population differences. We just have more suicidal (and murderous) antivaxxers.
No, this is flat out wrong. Perspective matters in terms of what you are measuring, and that is the wrong measure. Your example only makes sense if you are discussing the rate of adverse events. In that case, the incidence would have tripled if that's what is being measured. But that's an abuse of stats to say that something that causes 99% reduction in symptoms is three times as great as one that causes a 97% reduction. Factors of reduction is what is measured, which is what we're analyzing in the U.S. vs. Switzerland discussion.
Take the vaccine data. If I remember correctly, the original numbers of the Pfizer vaccine were something 95% effective, and Moderna was around 90% effective. In that moment, no one was calling Pfizer twice as effective of a vaccine, and for good reason.
Our interpretations have been different of the data in the past, but you've had my respect because you are generally well-informed. But there's nothing to argue about here if you are going to try to use sophistry to manipulate established mathematical principles.
The Moderna case proves my point, actually. People may have ignored the 90 vs 95% difference but the Moderna vaccine has proved to be substantially better! That 90 vs 95% signal was NOT noise!
Are you seriously going to argue against my cardiac drug example? 3x event rate is 3x the event rate. Even if both drugs work quite well!
I also forgot to reiterate that the reported death rate for US vs Switzerland broken out by vaxxed/unvaxxed/boosted is remarkably similar. This suggests that the US isn’t suffering from higher mortality across the board due to health differences. the US just has more antivaxxers and they are committed to dying for politics.
Yes, they are a moron. The math is abundantly clear for a healthy 20 year old. It's order of magnitude difference in risk for COVID vs the vaccine. There are reams of data showing this. It's long past the time for reasonable uncertainty. I am declaring them morons, and I am correct.
You're a moron if you think the data exists to make that assumption...
There's about 5,000,000 20 year old's in the US.
9.000 20 year old's died in 2020. That's a .18% chance of dying for any reason.
236 of them died with COVID. That's a .00472% chance of dying of COVID.
Not being worried about something that has a .00472% chance of happening doesn't make someone a moron. If they are healthy and not overweight the chance is an even tinier fraction of that .00472% chance.
Insisting that it does makes YOU a moron.
You're correct in the assessment that death from Covid shouldn't be a major concern for young people. However, it seems you think that young people ARE justified in being very concerned about potential vaccine side effects. The incidence of myocarditis as a vaccine side effect in 18-24 is in the range of .005-.01%. Why should a .01% risk of a hospital stay be so much more worrisome than a .005% risk of DEATH? And of course the risk of myocarditis from a Covid infection is roughly 10x HIGHER than as a side effect from the vaccine. The data is very clear that vaccination is the correct health decision for the overwhelming majority of young people. Unfortunately, it's still a very small benefit and young people are likely to be ok no matter what they do. The main benefit of getting younger people vaccinated is the reduction in overall transmission. The vaccines were great at that against alpha and pretty marginal against Omicron, very marginal unless you're boosted. Still, 30-50% additional prevention against transmission is huge at the societal level in preventing unnecessary deaths, hospitalizations, billions in healthcare costs that we'll all eventually pay for, and all the other ancillary impacts. You need near total buy in to get those benefits though. Unfortunately, that's not possible in the US. Probably not even worth arguing about anymore.
Are you seriously going to argue against my cardiac drug example? 3x event rate is 3x the event rate. Even if both drugs work quite well!
Yes. The if the drug was given, say, 10^6 at risk people with the purpose to reduce heart attacks, and 999,997 didn't have heart attacks, then the drug had a successful outcome rate of 0.999997, which is insignificantly different than 0.999999. You measure what the drug is designed to do or the effect that it is intended to have.
If you are looking at side-effects of the drug, which in this case explicitly means something that was not expected to occur except for the fact that the drug was administered, then we would be talking out 3E-6 vs. 1E-6. But since we aren't, then that's an incorrect application of stats.
Also, to your comment on the vaccines, Moderna was initially fairing worse, but now is doing better, so the fact that it appeared to be inferior was either noise or it simply just had better long-term effects that weren't understood at the time of release.
Are you seriously going to argue against my cardiac drug example? 3x event rate is 3x the event rate. Even if both drugs work quite well!
Yes. The if the drug was given, say, 10^6 at risk people with the purpose to reduce heart attacks, and 999,997 didn't have heart attacks, then the drug had a successful outcome rate of 0.999997, which is insignificantly different than 0.999999. You measure what the drug is designed to do or the effect that it is intended to have.
It's true that if the effect sizes are both large enough, then any differences are likely noise. That is not the case here but props for constructing a pointless counterexample of low statistical power -- vaccine data from 1st world countries does not have this issue.
You do realize that drug trials use my "statistics sophistry" to read out their primary endpoints? You don't see "fraction of heart attacks prevented" on the top line you see "number of cardiac events." In reality, these measures are one and the same and their interpretation is subject to the goals of the study, but in no way is my analysis invalid.
You seem to think that because 3x and 4-5x are close on the number line, that there isn't a statistically significant difference between them. A lot of data has been shown to you that the US has lost people to COVID at 250% the rate of Switzerland. There is no doubt that Switzerland is faring better.
As I pointed out several time and you repeatedly ignore, if you break out death rates by vaccination status, they are nearly exactly the same between Switzerland and the US. There is not an across-the-board mortality burden in the US attributable to health differences. The US simply has more antivaxxers who are dying.
It's always been the antivaxxers. The data are abundantly clear.
Are you seriously going to argue against my cardiac drug example? 3x event rate is 3x the event rate. Even if both drugs work quite well!
Yes. The if the drug was given, say, 10^6 at risk people with the purpose to reduce heart attacks, and 999,997 didn't have heart attacks, then the drug had a successful outcome rate of 0.999997, which is insignificantly different than 0.999999. You measure what the drug is designed to do or the effect that it is intended to have.
If you are looking at side-effects of the drug, which in this case explicitly means something that was not expected to occur except for the fact that the drug was administered, then we would be talking out 3E-6 vs. 1E-6. But since we aren't, then that's an incorrect application of stats.
Also, to your comment on the vaccines, Moderna was initially fairing worse, but now is doing better, so the fact that it appeared to be inferior was either noise or it simply just had better long-term effects that weren't understood at the time of release.
To put it a different way, let's say a team goes 97-3 one year. Their coach retires and the new coach the team goes 99-1. You wouldn't say they did three times better, because you are comparing 0.97 and 0.99, not 0.03 and 0.01.
Yes. The if the drug was given, say, 10^6 at risk people with the purpose to reduce heart attacks, and 999,997 didn't have heart attacks, then the drug had a successful outcome rate of 0.999997, which is insignificantly different than 0.999999. You measure what the drug is designed to do or the effect that it is intended to have.
If you are looking at side-effects of the drug, which in this case explicitly means something that was not expected to occur except for the fact that the drug was administered, then we would be talking out 3E-6 vs. 1E-6. But since we aren't, then that's an incorrect application of stats.
Also, to your comment on the vaccines, Moderna was initially fairing worse, but now is doing better, so the fact that it appeared to be inferior was either noise or it simply just had better long-term effects that weren't understood at the time of release.
To put it a different way, let's say a team goes 97-3 one year. Their coach retires and the new coach the team goes 99-1. You wouldn't say they did three times better, because you are comparing 0.97 and 0.99, not 0.03 and 0.01.
Two issues:
1. It depends what you care about! Certainly the new coach lost 3 times less. For something like.. dying of COVID... we tend to tally negative events, not positive events, so it's reasonable to compare death rates instead of not-death rates.
2. With enough statistical power, you can, in fact, separate 0.01 from 0.03.
The gigantic datasets from national public heath departments do have this power, but.. it doesn't even matter! We are actually comparing 0.33 to 0.22 or so which is much less statistically demanding.
If you really believe that it is insignificant that the US has lost people to COVID at 250% the rate of Switzerland since vaccines rolled out, despite similar death rates by vaccine status, be my guest. I'll stick with the shockingly simple result of: more vax = less death.
Yes. The if the drug was given, say, 10^6 at risk people with the purpose to reduce heart attacks, and 999,997 didn't have heart attacks, then the drug had a successful outcome rate of 0.999997, which is insignificantly different than 0.999999. You measure what the drug is designed to do or the effect that it is intended to have.
It's true that if the effect sizes are both large enough, then any differences are likely noise. That is not the case here but props for constructing a pointless counterexample of low statistical power -- vaccine data from 1st world countries does not have this issue.
You do realize that drug trials use my "statistics sophistry" to read out their primary endpoints? You don't see "fraction of heart attacks prevented" on the top line you see "number of cardiac events." In reality, these measures are one and the same and their interpretation is subject to the goals of the study, but in no way is my analysis invalid.
You seem to think that because 3x and 4-5x are close on the number line, that there isn't a statistically significant difference between them. A lot of data has been shown to you that the US has lost people to COVID at 250% the rate of Switzerland. There is no doubt that Switzerland is faring better.
As I pointed out several time and you repeatedly ignore, if you break out death rates by vaccination status, they are nearly exactly the same between Switzerland and the US. There is not an across-the-board mortality burden in the US attributable to health differences. The US simply has more antivaxxers who are dying.
It's always been the antivaxxers. The data are abundantly clear.
Not ignoring, I'm choosing to focus on a clear statistical error. It's impossible to argue every minute point and stay on course. If you wouldn't double down on frankly a high school statistics mistake, we could have been debating the actual merits. If you'd just admit your mistake, maybe we could get back to areas that actually are grey (like how much the vaccine is affecting the disparate outcomes vs. obesity).
And yes, drug companies do use this measure. Every time you read a paper or hear a commercial that says "Drug X prevents three times as many migraines" they looking at the effectiveness of the drugs. It is the effect the drug is actually having. You'd only be right to measure cardiac events if it was, "How much does fried food increase your risk for heart attacks," because that's the effect of fried food.
There aren't more antivaxxer (defined loosely nowadays as anyone not getting the COVID vaccine) than in Switzerland because the same number of doses per capita has been administered. It may be targeted differently, but then there are more people of low risks groups that are "antivaxxers" in Switzerland. Impossible for it to be otherwise.
Not ignoring, I'm choosing to focus on a clear statistical error. It's impossible to argue every minute point and stay on course. If you wouldn't double down on frankly a high school statistics mistake, we could have been debating the actual merits. If you'd just admit your mistake, maybe we could get back to areas that actually are grey (like how much the vaccine is affecting the disparate outcomes vs. obesity).
And yes, drug companies do use this measure. Every time you read a paper or hear a commercial that says "Drug X prevents three times as many migraines" they looking at the effectiveness of the drugs. It is the effect the drug is actually having. You'd only be right to measure cardiac events if it was, "How much does fried food increase your risk for heart attacks," because that's the effect of fried food.
There aren't more antivaxxer (defined loosely nowadays as anyone not getting the COVID vaccine) than in Switzerland because the same number of doses per capita has been administered. It may be targeted differently, but then there are more people of low risks groups that are "antivaxxers" in Switzerland. Impossible for it to be otherwise.
It's not an error to say "Switzerland has a lower COVID death rate than the USA." You trying to claim otherwise is getting a little comical.
You are trying to say that a reduction from 1 to 0.33 is not materially different from a reduction from 1 to 0.2. Given the size of the datasets involved you would need to make a sophisticated statistical case that we don't have the power to detect such a difference, as it certainly seems that we do.
We are arguing past each other. You are correct that when the effect size is really large, differences in remaining events may simply be noise. This is highly context specific as I've given several examples that you ignore or dismiss with strawman examples. It is not a middle school math mistake to choose different endpoints or methods of data analysis to validate the intervention. You've given no real argument for why tallying COVID deaths before and after vaccination is a terribly inappropriate way to judge the efficacy of an intervention. See the graph from Monkeys:
You are trying to claim that there isn't much difference post-vaccine. There has been an enormous difference post vaccine. Your method of analysis masks this. Mine does not. Mine method is objectively better at highlighting the disparities between the two countries. You may not like this because it goes against your thesis, but it's true.
I realize I was reading the graph wrong for the US deaths by vaccine status. The AUC for the unvaxxed is much larger than for Switzerland. AUC for vaxxed is similar. This validates my hypothesis that it was the antivaxxers that caused the death disparities between the two countries in the post-vaccine era.
Considering the vax is by FAR the most effective way to prevent severe COVID outcomes and there are documented differences between SWI and USA (booster coverage, targeting, etc), I am confused why you would balk so much at the vax being the primary explainer. Dumb antivaxxers have repeatedly been the answer for most of the pandemic in the last year.
Not ignoring, I'm choosing to focus on a clear statistical error. It's impossible to argue every minute point and stay on course. If you wouldn't double down on frankly a high school statistics mistake, we could have been debating the actual merits. If you'd just admit your mistake, maybe we could get back to areas that actually are grey (like how much the vaccine is affecting the disparate outcomes vs. obesity).
And yes, drug companies do use this measure. Every time you read a paper or hear a commercial that says "Drug X prevents three times as many migraines" they looking at the effectiveness of the drugs. It is the effect the drug is actually having. You'd only be right to measure cardiac events if it was, "How much does fried food increase your risk for heart attacks," because that's the effect of fried food.
There aren't more antivaxxer (defined loosely nowadays as anyone not getting the COVID vaccine) than in Switzerland because the same number of doses per capita has been administered. It may be targeted differently, but then there are more people of low risks groups that are "antivaxxers" in Switzerland. Impossible for it to be otherwise.
It's not an error to say "Switzerland has a lower COVID death rate than the USA." You trying to claim otherwise is getting a little comical.
You are trying to say that a reduction from 1 to 0.33 is not materially different from a reduction from 1 to 0.2. Given the size of the datasets involved you would need to make a sophisticated statistical case that we don't have the power to detect such a difference, as it certainly seems that we do.
We are arguing past each other. You are correct that when the effect size is really large, differences in remaining events may simply be noise. This is highly context specific as I've given several examples that you ignore or dismiss with strawman examples. It is not a middle school math mistake to choose different endpoints or methods of data analysis to validate the intervention. You've given no real argument for why tallying COVID deaths before and after vaccination is a terribly inappropriate way to judge the efficacy of an intervention. See the graph from Monkeys:
You are trying to claim that there isn't much difference post-vaccine. There has been an enormous difference post vaccine. Your method of analysis masks this. Mine does not. Mine method is objectively better at highlighting the disparities between the two countries. You may not like this because it goes against your thesis, but it's true.
I realize I was reading the graph wrong for the US deaths by vaccine status. The AUC for the unvaxxed is much larger than for Switzerland. AUC for vaxxed is similar. This validates my hypothesis that it was the antivaxxers that caused the death disparities between the two countries in the post-vaccine era.
Considering the vax is by FAR the most effective way to prevent severe COVID outcomes and there are documented differences between SWI and USA (booster coverage, targeting, etc), I am confused why you would balk so much at the vax being the primary explainer. Dumb antivaxxers have repeatedly been the answer for most of the pandemic in the last year.
-My argument has always been that the death rate has been lower in Switzerland (despite being given the same # of vaccine doses), so I have no idea where you are getting that from. Your opening line is a part of my thesis.
-Because the vaccination rate is the same, but the outcomes are different, then no, it can't simply be that there are more antivaxxers in the U.S. They are as vaccinated of a country as we are.
-It's possible that targeted vaccines are a part of the difference. If that has happened, then what that indicates is that vaccinated young and healthy people isn't necessary and we should focus on the vulnerable and the elderly. Because clearly if the same number of doses are given, and more of one group is getting them, less of another group is getting them. I have no problem with that if that is the case.
-I've only argued one point consistently for the sake of focus, which is purely mathematical in nature the past few threads. On your interpretation of percent decrease, yes, the mistake was a remedial math error, but I honestly don't care that it happened because we all screw up from time to time. It's the attempt to gaslight and pretend that this is in fact how stats are supposed to be done is what has me riled up. There are other points you've made that have value. Sometimes I've agreed with you in the past. Perhaps I'm starting to realize I should be more skeptical of your interpretations given that you won't let it go when you've clearly erred on interpreting percent decrease.
-The above point isn't even talking about the Switzerland vaccine effectiveness anymore since we can't agree on basic math. Your graphs look like there may be something there, but I'm honestly not seeing the point of debating a more nuanced and complex phenomenon if we are having a "Does 2+2=4? conversation."
-I've only argued one point consistently for the sake of focus, which is purely mathematical in nature the past few threads. On your interpretation of percent decrease, yes, the mistake was a remedial math error, but I honestly don't care that it happened because we all screw up from time to time. It's the attempt to gaslight and pretend that this is in fact how stats are supposed to be done is what has me riled up. There are other points you've made that have value. Sometimes I've agreed with you in the past. Perhaps I'm starting to realize I should be more skeptical of your interpretations given that you won't let it go when you've clearly erred on interpreting percent decrease.
-The above point isn't even talking about the Switzerland vaccine effectiveness anymore since we can't agree on basic math. Your graphs look like there may be something there, but I'm honestly not seeing the point of debating a more nuanced and complex phenomenon if we are having a "Does 2+2=4? conversation."
-Good day.
It's not gaslighting to have different ways of analyzing the same data.
Your original argument was:
Both countries had significant decreases in deaths. If you look at last winter vs this winter, Switzerland went down by a factor of 4-5 times, and the U.S. went down by a factor of about 3 times. Small difference,
Regardless of the definition of 'small difference' this analysis ignores or discounts the dramatic divergence in COVID deaths post-vaccine. There is no disputing that. COVID death rate disparity between Switzerland and USA has increased since vaccination started. The difference between 3x and 4.5x decrease has resulted in very real and significant mortality differences. That basically the opposite of what you are claiming, I believe.
Also I think we have to agree to disagree here:
If you only compared 100/100 = 1 and 100/33 .333=3 and claimed that the first treatment was 300% better than the other, that'd be the stuff of crackpots.
You absolutely can claim that the first treatment results in 3-fold less events assuming your trials were appropriately powered to detect such a difference! That makes the drug better!
Perhaps we just disagree because you think there isn't a statistically significant difference between 3x and 4.5x decrease and I do. I think that the quality of the data and the cumulative death data support that, yes, post-vaccine Switzerland has done far better than the US. The divergence post-vaccine is more evidence that background health factors are not a primary explanatory role.
I never thought were were having a 2+2=4 conversation, FWIW. We clearly disagreed on the appropriate way to analyze the data and I was deeply offended and/or flattered by someone deploying my own 'middle school math' burn on myself.
"“It’s really, massively stupid not to welcome vaccination,” the 98-year-old Munger said. “We probably have 30% of the people in the country that think vaccination is evil and [are] coming after them like the hobgoblins. It’s not good that there’s that much ignorance left.”"
"“It’s really, massively stupid not to welcome vaccination,” the 98-year-old Munger said. “We probably have 30% of the people in the country that think vaccination is evil and [are] coming after them like the hobgoblins. It’s not good that there’s that much ignorance left.”"
He appears quite out of touch.
The vaxx does not stop transmission. We will all get, or have had omicron. Even Fauci had admitted this. Vaxxed or not. The vaxx carries risks, some known, others not yet determined. The more jabs you take, the more you compromise your immune system. Omicron is like having a head cold. The only possible reason to take the jab is to lessen your symptoms (if that's true -- who knows?) -- from a head cold. No reason whatsoever to mandate vaccines or vaccine passports. By the way, when did natural immunity disappear from the planet.
It’s absurd, but expected, anti vaxxer stupidity to assert that 98% vs 99% relative risk reduction is meaningless or immeasurable.
These kind of hand-wavy “differences in small numbers don’t matter” are the ultimate last refuge when they cannot ignore the data anymore. Reminds me of DanM claiming 800k deaths was only 0.2% the US population so “who cares.”
We clearly disagreed on the appropriate way to analyze the data and I was deeply offended and/or flattered by someone deploying my own 'middle school math' burn on myself.
OK that cracked me up, so +1 for maintaining a good sense of humor.
I'm not trying to be a jerk here and have long maintained that the best way to convince someone to disagree with you is by being one. I've made my point and about the math part and pretty sure it stands for itself, so there's no benefit in continuing to bang my head against the wall.
"US death rate from covid is much higher than other wealthy nations"
Is this even the right comparison to make? Is the US really one of the "wealthy nations"? It is more accurate to say the US has the illusion of being a wealthy nation: a miniscule percentage of ultra-rich skew the numbers so that the US only appears to be comparable with Western Europe, Canada, and Japan. Most working-class Americans live a lifestyle comparable to Eastern Europe, NOT Denmark. For African-Americans, life is more comparable to a South American country.
The point being: the US covid rates should be compared to countries that are true peers, like Brazil or Poland (who probably still have lower rates than the US anyway, I'm guessing).
"“It’s really, massively stupid not to welcome vaccination,” the 98-year-old Munger said. “We probably have 30% of the people in the country that think vaccination is evil and [are] coming after them like the hobgoblins. It’s not good that there’s that much ignorance left.”"
Charlie is 0 for 3 so far this year. THIS is what's really bothering him:
We clearly disagreed on the appropriate way to analyze the data and I was deeply offended and/or flattered by someone deploying my own 'middle school math' burn on myself.
OK that cracked me up, so +1 for maintaining a good sense of humor.
I'm not trying to be a jerk here and have long maintained that the best way to convince someone to disagree with you is by being one. I've made my point and about the math part and pretty sure it stands for itself, so there's no benefit in continuing to bang my head against the wall.
Cheers.
Cheers. Apologies for any acrimony. The majority of LRC posters bring out (and deserve) my bad side but you should not.
