I believe I have done this in essence, perhaps a bit simplistic, but in keeping with the article - "an approximate answer to the right question".
Due to time constraints I've ignored the extreme cases (0.05 and 0.95), but have done the others, and the extreme cases are clearly not necessary as the actuals are closer to the other cases.
Obviously there may be confounding factors, but the conclusions seem straightforward - I won't spell them out for you.
I haven't exported 4 simulations for each scenario, as they all look very similar, but that's easy to do if you want to just run the code multiple times.
As requested, I've uploaded the code, data and images to github for you to review:
To run the code you'll need to download and extract the data, and then point the first few lines of the script "RVER.R" to the extracted files.
This is a best case scenario for the vaccines in that I've based the simulated distributions off the actual distribution. If they were fully randomized, they would look even more different.
I note that there are a few hundred counties where it appears that there are 0% vaccinated. If that is accurate (I haven't attempted to verify the integrity of the data), then we could use that as a control group (in a separate exercise), and I suspect things would look even less favorable for the vaccines.
The number of cases (positive test results) is important for people who still believe that the vaccination campaign will finally stop the virus from spreading by bringing R0 to zero. We know the vaccines cannot do that (too many breakthrough cases), instead the spread will stop when enough people got natural immunity (with or without vaccination).
So for us, another number is interesting: How many people will die or go into the ICU until herd immunity is reached? Vaccines promise to lower that number, but this effect seems to get weaker over time (and, of course, ivermectin is much better in reducing the CFR).
The scatterplot got me curious about what one would look like for all cause mortality. I'm not great with data analysis, so I probably set this up incorrectly (comments/feedback/independent verification most encouraged).
2) I scraped the statewide data from us mortality here and did a simple percent YoY change in deaths 2021 vs 2020 for week 35. https://www.usmortality.com/excess. I used week 35 because it looked like more recent data isn't complete. I broke these out by the age cohorts available. I realized after making the charts I probably should have used week 35 vax rates? Anyhow, the pictures are very similar if I do, so I didn't redo them.
3) did the x/y scatter plots in excel with a trendline and R2
Observations:
- in general, deaths are higher than 2020
- the relationship, such as it is, is backwards for the most vulnerable cohort, but it's an R2 of basically 0
- I don't remember a whole lot from undergrad stats, but R2 values of <0.04 aren't very impressive
Where do you guys get all this data and what tools do you use to run the plots? I need to start brushing up on some data science. If only to learn to think link one and understand the process a bit more.
I have the links to the data I used in the post you replied to. The CDC lets you download an excel sheet directly. For the US mortality data, I just wrote a script and pulled down their data using the REST API their webpage uses for various filtered links and saved it into a CSV file.
How much of this signal we're trying to tease out is entirely obfuscated by shitty testing protocols? I fear the pharma-devoted crowd will fall back on, "Well, public testing with bad implementation and design is creating sufficient false positives to obscure VE."
In order to prove vaccine effectiveness at "stopping the spread", they should be able to demonstrate lower case rates amongst the vaccinated. But, that's going to be hard to do with something as sensitive as a PCR test and the epidemic of testing they've created (due to mandates, neurosis, etc).
The phrase "hoisted by their own petards" comes to mind...
I love this idea of utilizing your readers! I can't wait to see what someone produces.
Hi Matthew,
I believe I have done this in essence, perhaps a bit simplistic, but in keeping with the article - "an approximate answer to the right question".
Due to time constraints I've ignored the extreme cases (0.05 and 0.95), but have done the others, and the extreme cases are clearly not necessary as the actuals are closer to the other cases.
Obviously there may be confounding factors, but the conclusions seem straightforward - I won't spell them out for you.
I haven't exported 4 simulations for each scenario, as they all look very similar, but that's easy to do if you want to just run the code multiple times.
As requested, I've uploaded the code, data and images to github for you to review:
https://github.com/DrMikeJS/RandomizedVaccineEfficacy .
To run the code you'll need to download and extract the data, and then point the first few lines of the script "RVER.R" to the extracted files.
This is a best case scenario for the vaccines in that I've based the simulated distributions off the actual distribution. If they were fully randomized, they would look even more different.
I note that there are a few hundred counties where it appears that there are 0% vaccinated. If that is accurate (I haven't attempted to verify the integrity of the data), then we could use that as a control group (in a separate exercise), and I suspect things would look even less favorable for the vaccines.
Anyway, best regards and keep up the good work!
Michael
Your displays are quite elegant. Thank you. I will be in touch.
The number of cases (positive test results) is important for people who still believe that the vaccination campaign will finally stop the virus from spreading by bringing R0 to zero. We know the vaccines cannot do that (too many breakthrough cases), instead the spread will stop when enough people got natural immunity (with or without vaccination).
So for us, another number is interesting: How many people will die or go into the ICU until herd immunity is reached? Vaccines promise to lower that number, but this effect seems to get weaker over time (and, of course, ivermectin is much better in reducing the CFR).
Do you mind linking to the data?
I think it's here:
https://data.cdc.gov/Vaccinations/COVID-19-Vaccinations-in-the-United-States-Jurisdi/unsk-b7fc
Thanks
Dear Crawford,
Can you provide an address for sending the project?
Preliminary results can be found at https://ibb.co/album/BHsNCK. The results not intended for making any conclusions, caveats apply.
The scatterplot got me curious about what one would look like for all cause mortality. I'm not great with data analysis, so I probably set this up incorrectly (comments/feedback/independent verification most encouraged).
https://imgur.com/a/IjBzD4r
1) the vax rate is the latest statewide cdc data from here, Series_Complete_Pop_Pct
: https://data.cdc.gov/Vaccinations/COVID-19-Vaccinations-in-the-United-States-Jurisdi/unsk-b7fc. I didn't break this out by age cohorts because I couldn't find that information. Plus, part of the jabbers argument is that a 45 year old needs to get jabbed to save grandma, so I'm not sure it would be appropriate.
2) I scraped the statewide data from us mortality here and did a simple percent YoY change in deaths 2021 vs 2020 for week 35. https://www.usmortality.com/excess. I used week 35 because it looked like more recent data isn't complete. I broke these out by the age cohorts available. I realized after making the charts I probably should have used week 35 vax rates? Anyhow, the pictures are very similar if I do, so I didn't redo them.
3) did the x/y scatter plots in excel with a trendline and R2
Observations:
- in general, deaths are higher than 2020
- the relationship, such as it is, is backwards for the most vulnerable cohort, but it's an R2 of basically 0
- I don't remember a whole lot from undergrad stats, but R2 values of <0.04 aren't very impressive
Criticisms?
that should have said us mortality here (usmortality.com)
Where do you guys get all this data and what tools do you use to run the plots? I need to start brushing up on some data science. If only to learn to think link one and understand the process a bit more.
I used Microsoft Excel for the plots.
I have the links to the data I used in the post you replied to. The CDC lets you download an excel sheet directly. For the US mortality data, I just wrote a script and pulled down their data using the REST API their webpage uses for various filtered links and saved it into a CSV file.
How much of this signal we're trying to tease out is entirely obfuscated by shitty testing protocols? I fear the pharma-devoted crowd will fall back on, "Well, public testing with bad implementation and design is creating sufficient false positives to obscure VE."
You'll get the joke after you see the plots. :-)
In order to prove vaccine effectiveness at "stopping the spread", they should be able to demonstrate lower case rates amongst the vaccinated. But, that's going to be hard to do with something as sensitive as a PCR test and the epidemic of testing they've created (due to mandates, neurosis, etc).
The phrase "hoisted by their own petards" comes to mind...