Coronavirus Outbreak

[nett]

Boot, Igor goes into deep detail about how he performed the linear regression on the German data. you can see it here: https://igorchudov.substack.com/p/covid ... ng-germans

Do you see any issues with his statistics? They seem pretty straight forward to me. You can look through the comments if you want to see age-stratified analysis, which was found to have no impact on excess death (standardized against 2011 data)

I know this exists for his analysis of the UK data too, but I don't have time to look for it right now.

[Bootstrap]

I think Nett's major claim is that the vaccines are more likely to kill you than the vaccine, that anyone with even a basic knowledge of statistics can see that. When I look at the same data he says he is looking at, for England and Wales in the same time period, it doesn't seem to show what he says it shows.

Whoa. I did not make that claim, at least in the short term. You need a pretty solid basis in statistics to make sense of most of this.

OK, maybe I was just mistaken. Apologies if I was.

But now you seem to be saying that there is some deep knowledge of statistics that I apparently don't yet have that would show me that my analysis so far is wrong. Could you please share that knowledge with me, looking at the same dataset in England and Wales that you proposed? I showed two different analyses of that same data, by different groups, with and without age-normalization (which took me quite a while to figure out). I provided a Google Collab notebook that loads the data and gives you exactly the formulas to run to compare this with other things.

There are certainly people who know a WHOLE lot more about statistics than I do. I suspect the people who did the two analyses I presented know a lot more about statistics than I do. If there's something else I need to learn so that I can know that there's something wrong with their analysis, could you please explain what it is and teach me how to interpret it correctly?

What is it that I am missing?

[nett]

I think Nett's major claim is that the vaccines are more likely to kill you than the vaccine, that anyone with even a basic knowledge of statistics can see that. When I look at the same data he says he is looking at, for England and Wales in the same time period, it doesn't seem to show what he says it shows.

Whoa. I did not make that claim, at least in the short term. You need a pretty solid basis in statistics to make sense of most of this.

OK, maybe I was just mistaken. Apologies if I was.

But now you seem to be saying that there is some deep knowledge of statistics that I apparently don't yet have that would show me that my analysis so far is wrong. Could you please share that knowledge with me, looking at the same dataset in England and Wales that you proposed? I showed two different analyses of that same data, by different groups, with and without age-normalization (which took me quite a while to figure out). I provided a Google Collab notebook that loads the data and gives you exactly the formulas to run to compare this with other things.

There are certainly people who know a WHOLE lot more about statistics than I do. I suspect the people who did the two analyses I presented know a lot more about statistics than I do. If there's something else I need to learn so that I can know that there's something wrong with their analysis, could you please explain what it is and teach me how to interpret it correctly?

What is it that I am missing?

You need a pretty solid statistical background to understand why the COVID trials did not show efficacy, nor safety. I explained this in great detail to you months ago, you said you would look at it in detail, and you never responded, I don't intend to do that again, because it took me hours to make the post, which you just ignored.

What two analyses did you show? I saw the pymc modelling scenario, but I don't see any use in it, because excess deaths are estimated and published, and then reported, so don't need to be modeled. I guess their case is that excess death data is very noisy and laggy. I don't agree about the noise, and if it's laggy, that just makes the outcomes even more negative.

Igor does a breakdown of the raw UK data, and you clearly see excess deaths are very high, you don't need modelling, and you don't need fancy statistics to see this.

https://igorchudov.substack.com/p/exces ... se-than-it

[Bootstrap]

Whoa. I did not make that claim, at least in the short term. You need a pretty solid basis in statistics to make sense of most of this.

OK, maybe I was just mistaken. Apologies if I was.

But now you seem to be saying that there is some deep knowledge of statistics that I apparently don't yet have that would show me that my analysis so far is wrong. Could you please share that knowledge with me, looking at the same dataset in England and Wales that you proposed? I showed two different analyses of that same data, by different groups, with and without age-normalization (which took me quite a while to figure out). I provided a Google Collab notebook that loads the data and gives you exactly the formulas to run to compare this with other things.

There are certainly people who know a WHOLE lot more about statistics than I do. I suspect the people who did the two analyses I presented know a lot more about statistics than I do. If there's something else I need to learn so that I can know that there's something wrong with their analysis, could you please explain what it is and teach me how to interpret it correctly?

What is it that I am missing?

You need a pretty solid statistical background to understand why the COVID trials did not show efficacy, nor safety. I explained this in great detail to you months ago, you said you would look at it in detail, and you never responded, I don't intend to do that again, because it took me hours to make the post, which you just ignored.

What two analyses did you show?

Could you please provide a link to that? I have frankly forgotten where it is. It does sound like a relevant topic.

I spent hours on the analysis in this thread too. So if you don't even know what two analyses I showed, I suggest that you go back and read this thread a little more carefully. I provided links, I quoted from the data of each, I answered questions you asked about each ...

This thread has discussed three separate analyses if England and Wales, using essentially the same data. Can you compare them? How do they differ? What are the statistical reasoning behind each?

I think I've done many hours of homework for you, laying things out so that someone with the statistical knowledge you claim should be able to do this much more efficiently than it would normally take. And presumably MUCH more efficiently than it took me.

[nett]

OK, maybe I was just mistaken. Apologies if I was.

But now you seem to be saying that there is some deep knowledge of statistics that I apparently don't yet have that would show me that my analysis so far is wrong. Could you please share that knowledge with me, looking at the same dataset in England and Wales that you proposed? I showed two different analyses of that same data, by different groups, with and without age-normalization (which took me quite a while to figure out). I provided a Google Collab notebook that loads the data and gives you exactly the formulas to run to compare this with other things.

There are certainly people who know a WHOLE lot more about statistics than I do. I suspect the people who did the two analyses I presented know a lot more about statistics than I do. If there's something else I need to learn so that I can know that there's something wrong with their analysis, could you please explain what it is and teach me how to interpret it correctly?

What is it that I am missing?

You need a pretty solid statistical background to understand why the COVID trials did not show efficacy, nor safety. I explained this in great detail to you months ago, you said you would look at it in detail, and you never responded, I don't intend to do that again, because it took me hours to make the post, which you just ignored.

What two analyses did you show?

Could you please provide a link to that? I have frankly forgotten where it is. It does sound like a relevant topic.

I spent hours on the analysis in this thread too. So if you don't even know what two analyses I showed, I suggest that you go back and read this thread a little more carefully. I provided links, I quoted from the data of each, I answered questions you asked about each ...

This thread has discussed three separate analyses if England and Wales, using essentially the same data. Can you compare them? How do they differ? What are the statistical reasoning behind each?

I think I've done many hours of homework for you, laying things out so that someone with the statistical knowledge you claim should be able to do this much more efficiently than it would normally take. And presumably MUCH more efficiently than it took me.

I'm not sure what else to say. take the data from the powerbi presentation and see what the excess data looks like. It's not really that hard. I went back through your posts, and all I see if the pymc.io post, which frankly makes no sense to me.

I do not intend to go back through 100s of posts to find that original discussion. Multiple people noticed it, and said something. if you want to change your reputation in this area, go find it yourself.

[Bootstrap]

I saw the pymc modelling scenario, but I don't see any use in it, because excess deaths are estimated and published, and then reported, so don't need to be modeled. I guess their case is that excess death data is very noisy and laggy. I don't agree about the noise, and if it's laggy, that just makes the outcomes even more negative.

OK, let me be blunt. I think you're bluffing. You have done nothing to convince me otherwise. I have repeatedly asked you to do the kinds of things that someone with the statistical background you claim should be able to do, some of them fairly easily. And now this ... seriously?

To me, this sounds like you might not know what statistical methods are for, how hypotheses are tested, what a statistical model is, etc. And you aren't doing what researchers do - you aren't providing a basis for your conclusions, you are just throwing a lot of strong adjectives around and claiming that you have superior statistical knowledge. If you do, please show it. I actually know all those adjectives too.

Let's start here: we are talking about "causal analysis". Suppose I say that Covid-19 is caused by Christmas trees. I can probably show a correlation between exposure to Christmas trees and Covid. And I can also show a bunch of excess deaths during the period after Christmas. But there might be other explanations, like suicide, depression, infectious disease. How do I look for the real cause? You typically need:

• Correlation. A and B do not correlate, A did not cause B.
• Sequence in time. If B happened before A, then A did not cause B. If there are cyclical patterns in B that do not correlate with the patterns of A, then A is an unlikely cause. etc.
• A plausible model for the proposed effect. Why would Christmas trees cause people to die? Why is that a plausible hypothesis?
• Eliminating the possibility of other clauses that might better explain the effect. At least the main culprits should be considered.

And we are in the field of "exploratory causal analysis". We can't really just do a controlled experiment. If I know that A is very likely the cause of a lot of excess deaths, I can't go out into shopping districts and A at people to see how many die. I have to look at data after the fact and create models that convincingly demonstrate that the correlations I am looking at are actually likely to be causal. If A is a proposed cure that is known to be safe, I can, of course, go A at a bunch of sick people. If A is a preventative measure that is known to be safe, I can, of course, go A a bunch of high risk people, such as seniors at nursing homes, and measure the results carefully in case A is less safe than I had thought. ( I use the word "I" very loosely, I am by no means THIS kind of researcher. )

So ... of course excess deaths need to be modeled. And you need a model for the proposed effect. And you need to propose a model for this effect and eliminate the possibility of other causes. This is on the level of "what is statistics" or "what is science".

Let's start here: what is an excess death?

Excess deaths are typically defined as the difference between the observed numbers of deaths in specific time periods and expected numbers of deaths in the same time periods. This visualization provides weekly estimates of excess deaths by the jurisdiction in which the death occurred. Weekly counts of deaths are compared with historical trends to determine whether the number of deaths is significantly higher than expected.

So ... how many deaths do you expect to happen next week? If more people die or less people die, why? Was the difference statistically significant or more likely to be random? Was it due to Covid-19 or vaccines or due to something else entirely? If it correlates with, say, Christmas, does that mean Christmas caused these deaths, or is there some common factor that might cause this correlation, such as cold weather, lots of people together in crowded spaces, people traveling and spending time together with people they are not usually with, etc?

FWIW, perhaps you could go back and look at Chudov's two analyses and grade them according to those four standards? Including the German one? You said you couldn't see anything wrong with them. Could you go back, using these four criteria, and see if there's something wrong?

[Bootstrap]

Could you please provide a link to that? I have frankly forgotten where it is. It does sound like a relevant topic.

!!! SNIP !!!

I do not intend to go back through 100s of posts to find that original discussion. Multiple people noticed it, and said something. if you want to change your reputation in this area, go find it yourself.

Stalemate, then. I can't find it, I don't even know what to search for, and I don't know if the claims you make about that thread are accurate or not. You don't want to find it.

If you want to me look at it and respond, or if you want evidence to attack my reputation with, perhaps you should try to find a link to that thread? Without it, there's nothing to discuss.

If you don't want to find the thread, maybe it's time to drop it?

[nett]

OK, let me be blunt. I think you're bluffing. You have done nothing to convince me otherwise. I have repeatedly asked you to do the kinds of things that someone with the statistical background you claim should be able to do, some of them fairly easily. And now this ... seriously?

To me, this sounds like you might not know what statistical methods are for, how hypotheses are tested, what a statistical model is, etc. And you aren't doing what researchers do - you aren't providing a basis for your conclusions, you are just throwing a lot of strong adjectives around and claiming that you have superior statistical knowledge. If you do, please show it. I actually know all those adjectives too.

ok, to be honest, you're just being a jerk at this point. Accusing me of bluffing is not a way to have a real conversation. I had typed up a much nastier reply, but I'm going to refrain from stooping to your level.

I have no interest in continuing this thread with you, here's a summary of why.

I made statements of two things I believe:

• Vaccines are not effective at preventing COVID, symptoms or death - your response was to post a graph from outworldindata, which is age standardized against 2013 inexplicable. I tried to start a conversation about how incentivized govt agencies are to misrepresent vaccine efficacy, but you simply ignored me. This would simply turn into a my data vs your data battle, which is pointless. No amount of data will change your mind here.
• Vaccine rollout it strongly correlated with excess death - your response to this was truly baffling, and I should have pointed that out and moved on, instead of engaging at all. You posted a pymc article about COVID-19 and excess death, which was not remotely relevant to our conversation, as it said nothing about vaccines as a confounding factors. I tried to redirect to Igor's straightfoward analysis of the data, but you refused to even acknowledge it, you just wanted to discuss this unrelated analysis. I'm not sure why you decided to demean and ridicule me over this either. The excess death model still makes no sense in the context of our conversation. Do you think causal inference is purely a statistical thing? Does this study you reference even mention confounding factors or attempt to correct them?

There is a common saying among people who don't do stats for a living: correlation does not equal causation. My stats professor advised to avoid people who say that, as well as those who do computer modeling, as they are generally unable to grasp the inherent unknowability of probabilities, and believe everything can be proven. I should have heeded his advice in this case.

Do you even know what the pymc article was trying to show? did you miss the end? What were you trying to show with it exactly?

I would like to reference it directly, but the page is completely broken at the moment:

[nett]

Could you please provide a link to that? I have frankly forgotten where it is. It does sound like a relevant topic.

!!! SNIP !!!

I do not intend to go back through 100s of posts to find that original discussion. Multiple people noticed it, and said something. if you want to change your reputation in this area, go find it yourself.

Stalemate, then. I can't find it, I don't even know what to search for, and I don't know if the claims you make about that thread are accurate or not. You don't want to find it.

If you want to me look at it and respond, or if you want evidence to attack my reputation with, perhaps you should try to find a link to that thread? Without it, there's nothing to discuss.

If you don't want to find the thread, maybe it's time to drop it?

Here it is. It's rather pointless now, because most countries have accepted natural immunity as being as strong as vaccinated, and it's mainstream knowledge that natural immunity lasts longer. However, at the time, you decided to just ignore me, while claiming I was making up my own claims

Like I said, these threads are pointless, you don't discuss in good faith, make nasty accusations, and when you finally run out of mainstream content to copy and paste, you just disappear.

Ok, I'll give it a shot.

First I did not distinguish between positive baseline N-binding antibody test and positive nucleic acid amplification test. Differentiating them does not result in enough incidents to be statistically significant.

here are the relevant incidences of reinfection among placebo and vaccinated populations from the 6 month study

16/716 placebo = 0.022
12/682 vaccinated = 0.018

subtracting we get an absolute risk reduction (ARR) of .004. this would not be statistically significant enough to show efficacy, but we can use it for comparison, because there were enough incidents for the population results to be significant.

We will now calculate Number Needed to Treat using the formula: NNT = 1/ARR

This gives us NNT = 250. meaning we need to vaccinate 250 patients with prior infection in order to prevent 1 infection via vaccination.

Meanwhile, the reactogenicity study found a Number Needed to Harm (NNH) of around 11 for moderate to severe reactions to the COVID vaccine. The data for that study is not presented as precisely, but this is what i was able to put together from the graphs, and it matches what other's analysis came up with.

Note that the reactogenicity study found a significantly increased risk of adverse reaction to the COVID mRNA vaccines in those with prior infection as compared to those without. I did not factor this in to my NNH. if i had, it would make the results even more extreme.

Consistently, compared to the first dose of the vaccine, we found an increased incidence and severity of self-reported side effects after the second dose, when recipients had been previously exposed to viral antigen. In view of the rapidly accumulating data demonstrating that COVID-19 survivors generally have adequate natural immunity for at least 6 months, it may be appropriate to re-evaluate the recommendation for immediate vaccination of this group.

Simply taking NNT / NNH from the two studies, you end up with 250 / 11 ≈ 23

please do note, that the reactogenicity study is not a pre-print, and was published in Life

[Bootstrap]

OK, let me be blunt. I think you're bluffing. You have done nothing to convince me otherwise. I have repeatedly asked you to do the kinds of things that someone with the statistical background you claim should be able to do, some of them fairly easily. And now this ... seriously?

To me, this sounds like you might not know what statistical methods are for, how hypotheses are tested, what a statistical model is, etc. And you aren't doing what researchers do - you aren't providing a basis for your conclusions, you are just throwing a lot of strong adjectives around and claiming that you have superior statistical knowledge. If you do, please show it. I actually know all those adjectives too.

ok, to be honest, you're just being a jerk at this point.

No, I'm refusing to be distracted from the basic questions about the data. Why all the insults and accusations and adjectives and boasts and changing the subject? No need to answer that, I don't want to speculate and I don't care, but none of this is teaching anybody anything at all about the data.

Accusing me of bluffing is not a way to have a real conversation.

Making tons of noise to distract from the actual content of the conversation is no way to have a real conversation. I'm describing your behavior. I don't know your motivation, I am not your judge, but the goal of expertise is to explain, to shed light on things, to explore together, to bring additional understanding, not to play King of the Mountain.

If you want to claim that expertise, please put it to good use. Respond to the content of what I say as though I actually said something. If you don't want to, there's no need for you to do so. But you are basically heckling. Why, I have no idea.

I have no interest in continuing this thread with you, here's a summary of why.

Your choice, feel free. I'm happy to talk to you about any subject at any time. But not this way.

I would like to discuss whatever subject we claim to be discussing. I don't care who has greater expertise, but I would like us to share what we know in a straightforward manner, answering questions or not, but without posturing.

I disagree with your summary of what has happened here, but that's fine, anyone who wants can go back and read the thread and draw their own conclusions.

Perhaps it would be better for us to focus on subjects like Scripture, where we seem to have more agreement.