Missed this, sorry. My comment below is largely an expanded version. I'm wondering how this was calculated (relative risk? odds ratio?). How the groups were made up. Etc. But want to give you credit here for making the key point first.
And 2 planes brought down 3 towers in a free fall. Fool me once shame on u= fool me twice shame on me. They think they can play the masses for a fool all the time! The ICI hearings begin today being totally reported on by MSM
The backstep dance is not an easy one for the liars to do. The moves involve: Let's get some other things in the top of the news, make believe what we did was a great success, reiterate that it would have been even better if everyone followed the official line, never admit to errors or past statements that are by now glaringingly obvious - - just ignore them and morph them into something else.... then repeat and repeat and repeat.
I'd ask as well: Was this a relative risk ratio? Or an odds ratio? Which are not the same. Our CDC Director has done this too often for my tastes -- given context-free comparisons, and not directly sourcing the data.
Also, as you point out, this is at best an observational study. We do not have two groups the same way we would in a RCT. So we need a clear and consistent definition of "unvaccinated" -- they keep changing this. We need clear and consistent definition of "boosted" -- one booster, two boosters. We need the time frame. We should further stratify the observed outcomes by age and biological sex. And we still have the huge problem of "dying with" as opposed to "dying from."
So if I have a small group of consisting of largely elderly or comorbid individuals in the "unvaccinated" pool, and I have a larger more representative population sample in the "booster" pool, I already know I have a higher percentages of deaths in Un_group as opposed to Bo_group. Right? Confounding and lurking variables. Were these blocked for and/or controlled.
This is also very time-frame sensitive. If I take the results for one month, I might get very difference outcomes than if I take the results for an entire year.
But let's keep going. Just crunch some pretend numbers In my Un_group (defined and selected however), I have a death percentage of say 12%. (Even without Covid, I'd have a death percentage well above the population average. Because much older people and sicker people are more likely to die). In my Bo_group (defined and selected however), I have a death percentage of 0.1. Still pretty high, given that healthy people under the age of 60 have 99.9-something probability of not dying from Covid.
Okay. Now let's calculate the relative risk ratio. 12 / 0.1 120 times greater! Shocking. Stunning! Horrific!
Maybe. Or, maybe not. Again, we need the Un_group and Bo_group to have large enough sample sizes, and to reasonably match up with each other. No dramatic differences in the age range distributions, the co-morbidity distributions, biological sex, etc. We need the data.
I am sorry to say that I have learned not to trust Dr. Wallensky's summaries. Data, and source, please.
A general comment about abuse of statistics, for example '…people who are boosted are 97 times less likely to die than the “unvaccinated”.'
Any time I see something about a ratio, or see the words 'relative risk', or something is 'X times more likely', my antennae start to quiver. I've been familiar with this for at least 12 years since I started looking at scientific studies in nutrition. But it's become glaring during the pandemic, if you pay attention, how general unfamiliarity with statistics is being taken advantage of.
Just yesterday I ran across something I had never heard of before, test-negative case control studies. As with relative risk, there may be a very specific purpose for this kind of thing, but when taken out of context it's incredibly misleading.
Right, I do believe the presentation of results by our CDC Director has been misleading -- even when what she said was not technically false. I've been upset with her on more than one occasion. Relative Risk itself is a standard comparison tool. If we are talking about the results from a Random Controlled Trial, the Relative Risk outcome is a highly significant (really strong) indicator. In some cases, call it proof.
But if we are talking about an observational study as we are here (which also btw shades into longitudinal study or retrospective study), then Relative Risk is still a highly useful and appropriate comparison. But not quite the knockdown-knock punch it would be for a RCT. More a useful and powerful indicator.
"If we are talking about the results from a Random Controlled Trial, the Relative Risk outcome is a highly significant (really strong) indicator. In some cases, call it proof."
Some cases. Take a hypothetical RCT. 100 people are treated, 100 people are controls. Two of the controls get sick. One of the treated people gets sick. This translates to a relative risk of 100%.
If you're evaluating the study in terms of your own personal risk, you need to look at the absolute risk which is 1% if you take the medicine and 2% if you don't. This may not even be statistically significant.
Closer to home, recently I looked at some infection fatality rates among hospitalized people who are vaccinated and unvaccinated. (The numbers are coming from my memory so might not be entirely accurate, but they're darn close.) Headlines were trumpeting 14 times as many unvaccinated die! Yeah, the unvaccinated fatality rate was .033% of hospitalized people dying; the vaccinated rate was .0024%.
People need to know what relative risk really is. And carefully examine what the numerators and denominators are, and pay attention to X-axes and Y-axes on all those pretty graphs we see every day.
Thanks for engaging on this. We're awash in statistics these days with very little understanding of how they're arrived at. Some good advice heard recently: pay attention to the data, but take any interpretation with a grain of salt.
Thank you. Important clarifications and corrections, indeed. And an excellent example about the fatality rate.
Let me also repeat your point -- one which I failed to make: for our own decision-making, we do need the absolute risk. Which might not be statistically significant.
That in part is what I was getting at with my unhappiness about Dr. Wallensky's context-free presentation. Relative risk ratios (and odds ratios) can have a magnifying effect. We might compare two very low probability outcomes. But still get a ratio where one appears considerably more likely than the other. 17 times greater. 60 times greater. Etc.
In truth, neither outcome is very likely. We have something that clinically does not matter. But we can do some math, and make the results look much impressive and scary.
Thank you again for your clarifications on this, and sharing your knowledge.
That's absolutely the reason for ignoring risk ratios when evaluating your own personal risk. Disingenuous deployment of relative risk is very common, especially in pharmaceutical studies >cough statins cough<.
In the IFR example I provided, you would compare those absolute risks versus the possibility of vaccine side effects, say. Or someone like me who doesn't take pharmaceuticals unless absolutely necessary – which is almost never, and I'm 68 – it helps to decide whether it's worth it or not.
Nearly half the deaths occurred before there ever was a vaccine, and the federal health authorities continue to designate a person as "unvaccinated" until two weeks AFTER the second injection. How many thumbs can you put on the scale?
Missed this, sorry. My comment below is largely an expanded version. I'm wondering how this was calculated (relative risk? odds ratio?). How the groups were made up. Etc. But want to give you credit here for making the key point first.
And 2 planes brought down 3 towers in a free fall. Fool me once shame on u= fool me twice shame on me. They think they can play the masses for a fool all the time! The ICI hearings begin today being totally reported on by MSM
The backstep dance is not an easy one for the liars to do. The moves involve: Let's get some other things in the top of the news, make believe what we did was a great success, reiterate that it would have been even better if everyone followed the official line, never admit to errors or past statements that are by now glaringingly obvious - - just ignore them and morph them into something else.... then repeat and repeat and repeat.
Will work for some but won't work for many.
I fact checked POTUS here on his tweet (sorry it's paywalled) but if anyone's interested!
https://nakedemperor.substack.com/p/fact-checking-the-us-president
I'd ask as well: Was this a relative risk ratio? Or an odds ratio? Which are not the same. Our CDC Director has done this too often for my tastes -- given context-free comparisons, and not directly sourcing the data.
Also, as you point out, this is at best an observational study. We do not have two groups the same way we would in a RCT. So we need a clear and consistent definition of "unvaccinated" -- they keep changing this. We need clear and consistent definition of "boosted" -- one booster, two boosters. We need the time frame. We should further stratify the observed outcomes by age and biological sex. And we still have the huge problem of "dying with" as opposed to "dying from."
So if I have a small group of consisting of largely elderly or comorbid individuals in the "unvaccinated" pool, and I have a larger more representative population sample in the "booster" pool, I already know I have a higher percentages of deaths in Un_group as opposed to Bo_group. Right? Confounding and lurking variables. Were these blocked for and/or controlled.
This is also very time-frame sensitive. If I take the results for one month, I might get very difference outcomes than if I take the results for an entire year.
But let's keep going. Just crunch some pretend numbers In my Un_group (defined and selected however), I have a death percentage of say 12%. (Even without Covid, I'd have a death percentage well above the population average. Because much older people and sicker people are more likely to die). In my Bo_group (defined and selected however), I have a death percentage of 0.1. Still pretty high, given that healthy people under the age of 60 have 99.9-something probability of not dying from Covid.
Okay. Now let's calculate the relative risk ratio. 12 / 0.1 120 times greater! Shocking. Stunning! Horrific!
Maybe. Or, maybe not. Again, we need the Un_group and Bo_group to have large enough sample sizes, and to reasonably match up with each other. No dramatic differences in the age range distributions, the co-morbidity distributions, biological sex, etc. We need the data.
I am sorry to say that I have learned not to trust Dr. Wallensky's summaries. Data, and source, please.
A general comment about abuse of statistics, for example '…people who are boosted are 97 times less likely to die than the “unvaccinated”.'
Any time I see something about a ratio, or see the words 'relative risk', or something is 'X times more likely', my antennae start to quiver. I've been familiar with this for at least 12 years since I started looking at scientific studies in nutrition. But it's become glaring during the pandemic, if you pay attention, how general unfamiliarity with statistics is being taken advantage of.
Just yesterday I ran across something I had never heard of before, test-negative case control studies. As with relative risk, there may be a very specific purpose for this kind of thing, but when taken out of context it's incredibly misleading.
https://jembendell.com/2021/12/23/lies-damn-lies-and-hospitalisation-statistics/
https://chrismasterjohnphd.com/covid-19/2022/02/04/test-negative-case-control-studies-are-a-scam
Right, I do believe the presentation of results by our CDC Director has been misleading -- even when what she said was not technically false. I've been upset with her on more than one occasion. Relative Risk itself is a standard comparison tool. If we are talking about the results from a Random Controlled Trial, the Relative Risk outcome is a highly significant (really strong) indicator. In some cases, call it proof.
But if we are talking about an observational study as we are here (which also btw shades into longitudinal study or retrospective study), then Relative Risk is still a highly useful and appropriate comparison. But not quite the knockdown-knock punch it would be for a RCT. More a useful and powerful indicator.
For a different comparison, I've did my own Relative Risk study based on CDC data. The outcomes were NOT flattering for our Covid policies in 2021: https://americanexile.substack.com/p/our-failed-covid-policies-now-a-womens
If you just want a better looking dashboard of the same: https://rpubs.com/Thom_JH/Covid_Relative_Risk
Thanks for raising the issue. & anyone who's actually read this, thank you for putting up with me.
"If we are talking about the results from a Random Controlled Trial, the Relative Risk outcome is a highly significant (really strong) indicator. In some cases, call it proof."
Some cases. Take a hypothetical RCT. 100 people are treated, 100 people are controls. Two of the controls get sick. One of the treated people gets sick. This translates to a relative risk of 100%.
If you're evaluating the study in terms of your own personal risk, you need to look at the absolute risk which is 1% if you take the medicine and 2% if you don't. This may not even be statistically significant.
Closer to home, recently I looked at some infection fatality rates among hospitalized people who are vaccinated and unvaccinated. (The numbers are coming from my memory so might not be entirely accurate, but they're darn close.) Headlines were trumpeting 14 times as many unvaccinated die! Yeah, the unvaccinated fatality rate was .033% of hospitalized people dying; the vaccinated rate was .0024%.
People need to know what relative risk really is. And carefully examine what the numerators and denominators are, and pay attention to X-axes and Y-axes on all those pretty graphs we see every day.
Thanks for engaging on this. We're awash in statistics these days with very little understanding of how they're arrived at. Some good advice heard recently: pay attention to the data, but take any interpretation with a grain of salt.
Thank you. Important clarifications and corrections, indeed. And an excellent example about the fatality rate.
Let me also repeat your point -- one which I failed to make: for our own decision-making, we do need the absolute risk. Which might not be statistically significant.
That in part is what I was getting at with my unhappiness about Dr. Wallensky's context-free presentation. Relative risk ratios (and odds ratios) can have a magnifying effect. We might compare two very low probability outcomes. But still get a ratio where one appears considerably more likely than the other. 17 times greater. 60 times greater. Etc.
In truth, neither outcome is very likely. We have something that clinically does not matter. But we can do some math, and make the results look much impressive and scary.
Thank you again for your clarifications on this, and sharing your knowledge.
That's absolutely the reason for ignoring risk ratios when evaluating your own personal risk. Disingenuous deployment of relative risk is very common, especially in pharmaceutical studies >cough statins cough<.
In the IFR example I provided, you would compare those absolute risks versus the possibility of vaccine side effects, say. Or someone like me who doesn't take pharmaceuticals unless absolutely necessary – which is almost never, and I'm 68 – it helps to decide whether it's worth it or not.
Professor Bendell did a great job running down the bullshit statistical method by owned pharma hacks, thank you.
Thanks, I was wondering where they got their data. 👍🏼
Nearly half the deaths occurred before there ever was a vaccine, and the federal health authorities continue to designate a person as "unvaccinated" until two weeks AFTER the second injection. How many thumbs can you put on the scale?
*And without comparison to unvaxxed-but-treated-early, it's not an accurate comparison ...
THIS!