Need Help With Upcoming Episode on Mask Junk Science

ok, but don't we need to go off of the headline data they presented:

View attachment 2171




here is the same data from the Stanford part of the team:

The numbers in that graphic come from "a regression of symptomatic seroprevalence on a treatment indicator, clustering at the village level and controlling for fixed effects for each pair of control-treatment villages." You would have to run backwards through the regression to get to the raw numbers, which isn't possible for us.

The researchers found that among the more than 350,000 people studied, the rate of people who reported symptoms of COVID-19, consented to blood collection and tested positive for the virus was 0.76% in the control villages and 0.68% in the intervention villages, showing an overall reduction in risk for symptomatic, confirmed infection of 9.3% in the intervention villages regardless of mask type.

unless they tell us otherwise I think we have to go with these numbers.

I agree. Which is why you have to start with "people who reported symptoms of COVID-19" and "consented to blood collection" (the numbers I gave earlier).
 
and as far as p value... if there were the exactly the same number of positive blood tests in both groups the p-value would be 1.

No. Even if the number of cases was the same in both groups, the number of people who weren't cases was quite different. Let's test this. I'll put the same number of cases in each group (I'll use the average), and do the Chi-square test...

Yup, p-value is 0.042.

ETA: The strength of the findings don't come from the number of cases - small changes make little to no difference. The strength comes from the large difference in non-cases between each group. And it would take a lot more work to erase that difference.
 
Last edited:
you have to start with "people who reported symptoms of COVID-19" and "consented to blood collection" (the numbers I gave earlier).


I wouldn't read:

the rate of people who reported symptoms of COVID-19, consented to blood collection and tested positive for the virus was 0.76% in the control villages

...that way. I mean it seems to me the most straightforward way to read it is ".76% tested positive for covid."

what makes you confident they're trying to say something differently?
 
ok, but don't we need to go off of the headline data they presented:

View attachment 2171




here is the same data from the Stanford part of the team:

The researchers found that among the more than 350,000 people studied, the rate of people who reported symptoms of COVID-19, consented to blood collection and tested positive for the virus was 0.76% in the control villages and 0.68% in the intervention villages, showing an overall reduction in risk for symptomatic, confirmed infection of 9.3% in the intervention villages regardless of mask type.

unless they tell us otherwise I think we have to go with these numbers.

and it looks to me like those numbers show a difference of about 9 cases between the control group and the intervention group.

View attachment 2173

Yes. That's what I think too. And there is no way that 9 is statistically significant.

I am hyper busy at work. I will give this the proper attention this afternoon and calculate P values for you.

I would like to see agreement from other members of the forum that these are the figures that should be used.
 
I wouldn't read:

the rate of people who reported symptoms of COVID-19, consented to blood collection and tested positive for the virus was 0.76% in the control villages

...that way. I mean it seems to me the most straightforward way to read it is ".76% tested positive for covid."

what makes you confident they're trying to say something differently?

Because I read the study and it says "among the 335,382 participants who completed symptom surveys, 27,166 (8.1%) reported experiencing COVID-like illnesses during the study period...omitting symptomatic participants who did not consent to blood collection, symptomatic seroprevalence was 0.76% in control villages and 0.68% in the intervention villages."
 
Because I read the study and it says "omitting symptomatic participants who did not consent to blood collection, symptomatic seroprevalence was 0.76% in control villages and 0.68% in the intervention villages."
I think they're just saying:

Over one-third (40.3%) of symptomatic participants agreed to blood collection. Omitting symptomatic participants who did
not consent to blood collection, symptomatic seroprevalence was 0.76% in control villages and
0.68% in the intervention villages. Because these numbers omit non-consenters, it is likely that
the true rates of symptomatic seroprevalence are substantially higher (perhaps by 2.5 times
, if
non-consenters have similar seroprevalence to consenters).


... but this seems to be a misuse of the term "rates." I suspect that the authors meant to say that the true number of "cases" are substantially higher
 
The numbers in that graphic come from "a regression of symptomatic seroprevalence on a treatment indicator, clustering at the village level and controlling for fixed effects for each pair of control-treatment villages." You would have to run backwards through the regression to get to the raw numbers, which isn't possible for us.
I agree. Which is why you have to start with "people who reported symptoms of COVID-19" and "consented to blood collection" (the numbers I gave earlier).

Peeking in quickly to see discussion progress.

Ellis, what I quote above, from you, is exactly what I said at some point yesterday and you replied that it would be terrible methodology and that's not what was done. In fact, you have doubled back a few times now to the same understanding that I originally stated on various points. I'm worried about you. We can all see what was written in the thread.

Now Alex is clearly as confused as I am as to what the study did and measured and where the significance came from. With all of your reversals, I suspect you are too, but won't admit it.

I will review very closely again this evening, but I am still convinced at this point that the study is, literally, gibberish - kind of like the emperor's new cloths. There's a mish mash of objectives, measurements, etc. that are supposed to sound meaningful, but are not. Buzz words to catch the attention of media and to show the WHO that the authors should be funded b/c they know how to play the game (arrive at a media hook that promotes the WHO's desired policy). Therefore, there is no way to understand what was done. We're not supposed to.

I am still waiting for you - Ellis - to spell out step by step, what the study did/measured and what the figures associated with each step are. Assume that the rest of us are morons and need to be enlightened. If you can do that, I'll even go get vaxxed and wear a mask ;-)

Come on, you spent enough time writing comments. You could have done what I'm asking for.
 
I think they're just saying:

Over one-third (40.3%) of symptomatic participants agreed to blood collection. Omitting symptomatic participants who did
not consent to blood collection, symptomatic seroprevalence was 0.76% in control villages and
0.68% in the intervention villages. Because these numbers omit non-consenters, it is likely that
the true rates of symptomatic seroprevalence are substantially higher (perhaps by 2.5 times
, if
non-consenters have similar seroprevalence to consenters).


... but this seems to be a misuse of the term "rates." I suspect that the authors meant to say that the true number of "cases" are substantially higher
At this point, I agree with your understanding, Alex. But there are still aspects of how they got there that are totally unclear to me.

Ellis had said they didn't measure blood serum of only symptomatic people. I said they did. I read what you quoted. Actually, I take Ellis point about why it would be stupid to blood test only symptomatic people.

I am still looking for what would be correct methodology - blood tested mask wearers and blood tested non-mask wearers. Blood test revealed, respectively X and Y number of positive blood tests in each group. But we won't find that because I don't think they did it. They talk like they did that, but then you don't see where they have those figures. Ellis is has said many contradictory things about that point. He has said they did it and he has said they did not. He has said they only tested people with self-reported symptoms and then used some arcane regression to arrive at some figure for non-self reported symptoms. This is not to pick on Ellis. Like I said, I think he is as confused as the rest of us. Just we admit it.

Worst experiment write up I have seen in a long time. Again, I think I know why.
 
Last edited:
Peeking in quickly to see discussion progress.

Ellis, what I quote above, from you, is exactly what I said at some point yesterday and you replied that it would be terrible methodology

Please show me where you think this happened.

In fact, you have doubled back a few times now to the same understanding that I originally stated on various points.

Please give me examples.

Thank you.
 
Please show me where you think this happened.



Please give me examples.

Thank you.
nope. not going there with you. It's all right in the thread (unless you've edited it out) anyone interested can look and see.

so when do we get to see your synopsis?
 
I think they're just saying:

Over one-third (40.3%) of symptomatic participants agreed to blood collection. Omitting symptomatic participants who did
not consent to blood collection, symptomatic seroprevalence was 0.76% in control villages and
0.68% in the intervention villages. Because these numbers omit non-consenters, it is likely that
the true rates of symptomatic seroprevalence are substantially higher (perhaps by 2.5 times
, if
non-consenters have similar seroprevalence to consenters).


... but this seems to be a misuse of the term "rates." I suspect that the authors meant to say that the true number of "cases" are substantially higher
I see what you're thinking. But if you add back in the non-consenters, assuming that they have the same COVID rate as the consenters, you are adding cases at a much greater rate than you are adding non-cases compared to the overall population (22% vs. 0.76/0.68%) . So your overall "symptomatic seroprevalence rate" will increase. Do you need me to do the math?
 
nope. not going there with you. It's all right in the thread (unless you've edited it out) anyone interested can look and see.

Except I haven't done what you said - contradicted myself or said something was good methodology one time, then bad methodology another.

So there has been a misunderstanding - I misunderstood you or you misunderstood me. I can't clarify anything for you wrt a synopsis if we are misunderstanding each other from the get go. Examples would be very helpful to see where we've gone wrong.

I said this:

"The numbers in that graphic come from "a regression of symptomatic seroprevalence on a treatment indicator, clustering at the village level and controlling for fixed effects for each pair of control-treatment villages." You would have to run backwards through the regression to get to the raw numbers, which isn't possible for us."

You said:
"Ellis, what I quote above, from you, is exactly what I said at some point yesterday and you replied that it would be terrible methodology and that's not what was done."

I looked through what you said yesterday. I found something similar from you:
"So, partly answering my own question, if you're going to look at every town individually, then a regression analysis begins to make sense. However, if that's what they did, then they should most definitely show the town level data in a table in their write-up.

So why are they looking at each town separately? is that explained? Does it make sense to approach the analysis that way? I don't know. Just asking because I think it is absolutely critical."

My response to that was:
"That would be to much data to include in the write-up (remember, there were 600 villages). But that kind of data is sometimes made available as a supplementary download. I would like to see it, too.

I said it was a gross oversimplification. Maybe I should have called it a metaphor. But you want to make the best use of the data, without adding in any opportunities for error or bias. That a number of measurements came from the same village is information that is lost if you just pool all the data. And villages which are larger or at higher risk (so there are more cases), could have a disproportionate effect on the results. Regression equalizes everything so outliers or other strangeness doesn't have a disproportionate effect. And at the same time it maximizes the use of the available information."

So, as far as I can tell, I said the result was "a regression clustering at the village level". Yesterday you said "if you're going to look at every town individually, then a regression analysis begins to make sense." Which is similar. And then I agreed with you about including the data somehow and gave a bit more explanation about why regression would be useful.

How is any of that me saying it was "terrible methodology and that's not what was done"? Or were you referring to something else?
 
Last edited:
(Venting)

https://www.osha.gov/laws-regs/regulations/standardnumber/1910/1910.134AppA

I work in Gas Station construction. In order to enter a confined space and perform fiberglass repair that include sanding, you are required by OSHA perform fit testing which is where an OSHA rep checks to see that you know how to secure a RESPIRATOR (not a n95 mask) properly so that it utilizes the filtration function.

For minor/quick fiberglass repairs, sometimes we'll use regular masks N-95 prefferably.

Basic human sensibility tells us that unless your using a respirator, your going to be breathing a significant amount of fiberglass.


The face mask I wear at my cubicle is not stopping anything, except quantity of oxygen and harassments about mandates.

The complication of this debate on testing makes it very clear that we're not simplifying the simple parts.
 
Except I haven't done what you said - contradicted myself or said something was good methodology one time, then bad methodology another.

So there has been a misunderstanding - I misunderstood you or you misunderstood me. I can't clarify anything for you wrt a synopsis if we are misunderstanding each other from the get go. Examples would be very helpful to see where we've gone wrong.

I said this:

"The numbers in that graphic come from "a regression of symptomatic seroprevalence on a treatment indicator, clustering at the village level and controlling for fixed effects for each pair of control-treatment villages." You would have to run backwards through the regression to get to the raw numbers, which isn't possible for us."

You said:
"Ellis, what I quote above, from you, is exactly what I said at some point yesterday and you replied that it would be terrible methodology and that's not what was done."

I looked through what you said yesterday. I found something similar from you:
"So, partly answering my own question, if you're going to look at every town individually, then a regression analysis begins to make sense. However, if that's what they did, then they should most definitely show the town level data in a table in their write-up.

So why are they looking at each town separately? is that explained? Does it make sense to approach the analysis that way? I don't know. Just asking because I think it is absolutely critical."

My response to that was:
"That would be to much data to include in the write-up (remember, there were 600 villages). But that kind of data is sometimes made available as a supplementary download. I would like to see it, too.

I said it was a gross oversimplification. Maybe I should have called it a metaphor. But you want to make the best use of the data, without adding in any opportunities for error or bias. That a number of measurements came from the same village is information that is lost if you just pool all the data. And villages which are larger or at higher risk (so there are more cases), could have a disproportionate effect on the results. Regression equalizes everything so outliers or other strangeness doesn't have a disproportionate effect. And at the same time it maximizes the use of the available information."

So, as far as I can tell, I said the result was "a regression clustering at the village level". Yesterday you said "if you're going to look at every town individually, then a regression analysis begins to make sense." Which is similar. And then I agreed with you about including the data somehow and gave a bit more explanation about why regression would be useful.

How is any of that me saying it was "terrible methodology and that's not what was done"? Or were you referring to something else?
So...that was a bust. So I decided to go at it from the other direction. I went back and looked at what I complained about as "bad methodology" yesterday. And the only thing I complained about was the comparison you made - "I'm using masked/not masked and blood tested positive/negative." But that has zero to do with regression, so it can't be what you were referring to, either.

See why I need you to show me what you're talking about? I'm hopeless at this.
 
Basic human sensibility tells us that unless your using a respirator, your going to be breathing a significant amount of fiberglass.

The face mask I wear at my cubicle is not stopping anything, except quantity of oxygen and harassments about mandates.

Just to clarify. This study isn't about whether the face mask you wear at your cubicle is stopping anything. It's the other way around. If you're breathing out a significant amount of fiberglass (I know, that's probably not a real thing), is the face mask you wear at your cubicle stopping some of it?
 
I see what you're thinking. But if you add back in the non-consenters, assuming that they have the same COVID rate as the consenters, you are adding cases at a much greater rate than you are adding non-cases compared to the overall population (22% vs. 0.76/0.68%) . So your overall "symptomatic seroprevalence rate" will increase. Do you need me to do the math?

I don't see how/why "adding cases at a much greater rate" would matter in terms of .76%/.68%



1632338060444.png

still looks like 9 fewer cases to me.

I'd need some very clear and convincing "nail in the coffin" stuff from the study to conclude otherwise.
 
I don't see how/why "adding cases at a much greater rate" would matter in terms of .76%/.68%

It's 22% vs. 0.76% and 22% vs. 0.68%.

The math:

New case numbers are:

Mask 2.5 x 1131 = 2827
No-mask 2.5 x 1162 = 2905

New population numbers are line 2 from Table 1 without any omissions:

Mask 174,171
No-mask 161,211

New rates:

Mask 2827/174,171 x 100% = 1.6%
No-mask 2905/161,211 x 100% = 1.8%

New Chi-square p-value = 0.00007
 
Last edited:
Just to clarify. This study isn't about whether the face mask you wear at your cubicle is stopping anything. It's the other way around. If you're breathing out a significant amount of fiberglass (I know, that's probably not a real thing), is the face mask you wear at your cubicle stopping some of it?
Thank you for the explaining this simply.
I can answer it very simply while paraphrasing Fauci; "Maybe a droplet."
Other than a droplet, I can ABSOLUTELY confirm that whatever outward filtration is created by non-fitted face masks will be ineffectual due to the labored breathing needed to meet oxygen needs which shoots out of all sides of the mask.
I assume (with basic human sensibility) that the spread of exhaled air (other than droplets) is probably worse with masks due to the labored breathing.
The fact that for office/indoor settings, we're debating masking effectiveness and not ventilation / fresh-air requirements, tells you how far off the mark we are.

I argue that we can measure how serious a mask debate is by the amount of reference to and emphasis on fresh air circulation.

It's the difference between standing next to me in the ocean while peeing, vs sitting next to me in the hot tub while peeing.
How long would you stay in the hot tub?
 
Last edited:
Thank you for the explaining this simply.
I can answer it very simply while paraphrasing Fauci; "Maybe a droplet."
Other than a droplet, I can ABSOLUTELY confirm that whatever outward filtration is created by non-fitted face masks will be ineffectual due to the labored breathing needed to meet oxygen needs which shoots out of all sides of the mask.

You poor thing. That must be hard on you. Hopefully you're not the one breathing out fiberglass in this scenario.

It's the difference between standing next to me in the ocean while peeing, vs sitting next to me in the hot tub while peeing.
How long would you stay in the hot tub?

Realistically, it depends on if you tell me.
 
So...that was a bust. So I decided to go at it from the other direction. I went back and looked at what I complained about as "bad methodology" yesterday. And the only thing I complained about was the comparison you made - "I'm using masked/not masked and blood tested positive/negative." But that has zero to do with regression, so it can't be what you were referring to, either.

See why I need you to show me what you're talking about? I'm hopeless at this.
Just for the edification of others who might be following the thread, you said this yesterday,
"Using this measure, masks could prevent 99% of COVID cases, and your measure would still show up as statistically insignificant. Try it for yourself and see. Make up some huge number of people who report respiratory symptoms, in the no-mask group - let's say 10,000. Applying our seropositivity rate of 22%, that gives us 2200 in the no-mask/proven to have COVID group and 7800 in the no-mask/proven to not have COVID group. Now let's say only 100 people reported respiratory symptoms in the mask group. Applying our seropositivity rate of 22% gives us 22 in the mask/proven to have COVID group and 78 in the mask/proven not to have COVID group. Run your Chi-square test and see what you get...p=1. So masks prevented 2178/2200 cases, but your "outcome" shows that "masks don't make a difference"? What if it was the other way around. What if masks caused COVID and there were 2178 more cases in the mask group. But the researchers went ahead and claimed that masks weren't harmful, because p=1. Would you really be prepared to buy that?"
It's 22% vs. 0.76% and 22% vs. 0.68%.

The math:

New case numbers are:

Mask 2.5 x 1131 = 2827
No-mask 2.5 x 1162 = 2905

New population numbers are line 2 from Table 1 without any omissions:

Mask 174,171
No-mask 161,211

New rates:

Mask 2827/174,171 x 100% = 1.6%
No-mask 2905/161,211 x 100% = 1.8%

New Chi-square p-value = 0.00007

"assuming that they have the same COVID rate as the consenters" - big assumption. One I would not make when the numbers you are extrapolating from are, as Alex keeps correctly pointing out, so small.

I see now that this is the big flaw in the study. I wasn't getting it because it isn't something that can be done that way. There are all kinds of confidence intervals around that kind of extrapolation, again, especially when starting from such small numbers.

"New population numbers"? I don't like my results so let's create a new population and see if we get something I like.
 
Last edited:
Back
Top