The lack of full data transparency is all that matters. Period. (But because I'm curious and want to learn I've reproduced all the plots from various authors, which you can find below)
Kirsch wrote about John Sullivan's plot that "the key is the orange mortality rate goes from around 2,500 in August to over 5,000 in December".
However in 2018-2022, the CMR of the 80-89 age group in New Zealand ranged from about 6,635 to 7,419 deaths per 100,000 person-years, but the upper end of the age group is overrepresented in the NZ data compared to the lower end: https://mongol-fi.github.io/moar.html#Representation_of_age_groups_in_the_dataset. Based on the age composition of the vaccinated people, I got an average baseline of about 8,463 for the CMR in December 2021, which is higher than the historical CMR for ages 80-89 because ages 85-89 are overrepresented in the data compared to ages 80-84, and because I followed people who were 80-89 years old at the time of vaccination, so many of them were more than half a year older by December 2021.
When I took the first dose of each person who was 80-89 years old at the time of their dose and I followed the crude mortality rate of the cohort on each day until the end of the data, I got the same brief spike in mortality in June 2021 as Sullivan, but after that it took until June 2022 before the 14-day moving average of the CMR reached above the baseline. And even then the CMR only stayed above the baseline briefly, and June 2022 was in the middle of a COVID wave, and it was winter but I didn't account for seasonality when I calculated my baseline: https://mongol-fi.github.io/moar.html#Plot_by_John_Sullivan_for_crude_mortality_rate_by_date_in_ages_80_89.
The reason why the CMR of ages 80-89 was initially so far below the baseline could be because the healthy vaccinee effect seems to be stronger in older age groups than younger age groups.
You wrote that "The positive slope in 2023 could be due to the diminishing size of the cohort". You're probably right, because the baseline I calculated based on the age composition of the cohort also increased at a fairly steep rate in 2023, but another factor which produces the positive slope is that it was summer in early 2023 but winter in mid-2023.
In your analysis where you removed deaths in the July, August, and September of each year, you forgot to remove the person-days of alive people in summer months, which produced the two dents in mortality on either side of your "massive signal". When I reproduced your analysis, the mortality rate at the peak of the massive signal was still below the baseline based on the age composition, and when I also removed the person-days during summer months, I got rid of the two dents: https://i.ibb.co/my3zj81/massive-signal-dents.png.
I've quickly read through your very detailed analysis. Thanks for taking the time to document all that. I will try and reproduce UncleJohnReturns Simpson's paradox, that is indeed very interesting, (but it doesn't take away from the fact that many "fully" vaccinated elderly died during the Omicron wave in the summer of 2022 in New Zealand AND Australia. See https://kalev.substack.com/p/new-zealands-excess-deaths-a-case for New Zealand. And https://kalev.substack.com/p/australias-excess-deaths-a-case-study for Australia. Australia is particularly bad.)
> you forgot to remove the person-days of alive people in summer months
I'm not sure I follow. I completely removed all the people who died in July to September from the dataset and rebuilt the buckets from scratch and then looked at the MR. Since these people were removed from the dataset entirely they are not contributing to the alive person-day aggregates in summer months.
Yeah but the people who are alive in July-September are still added to the denominator of the mortality rate, which produces the dips in mortality rate during the weeks where winter months are overrepresented, like first around weeks 30 and 50 and a year later around weeks 80 and 100.
To avoid the bias, you could've just used the regular output of buckets.py but you could've removed all lines for July to September months, like I did in my second plot at IBB.
> To avoid the bias, you could've just used the regular output of buckets.py but you could've removed all lines for July to September months
That would introduce a bias. If you remove the July to September buckets after including everyone, the people who died in July to September would have contributed to the "alive" tally up until their death.
You can't count them as "alive" up until their death and then ignore their death.
Instead, what I did is completely remove these people from the analysis altogether, as if they never existed. As if they were never part of the dataset to begin with. They didn't contribute to any "alive" or "death" tallies.
(We are already dealing with an incomplete dataset. My understanding of Steve's approach is that an incomplete dataset does not matter. So if we remove even more records, it shouldn't matter. I did the same analysis for other age groups but it was only in the elderly where you saw an increase in MR after week 40.)
Let's not forget that the crux of the matter, and the whole reason for me writing this article, is that the total lack of transparency from many governments around the world is the real problem.
We shouldn't be left with incomplete datasets to try and draw conclusions.
As I mentioned in the article, if the evidence was overwhelmingly supportive of "vaccine safety and efficacy", they would relish demonstrating that to us. The fact that they aren't speaks volumes.
Agreed, transparency from government officials is a must.
The last few years have seen a huge increase in the lies, obfuscation, and misinformation by our “public servants” who are supposedly working for us but are if fact working for themselves and their leftist political agendas.
Great job. I totally agree. The lack of transparency tells you everything you need to know. I just emailed you give me a call and I’ll show you some things that nobody has seen yet.
Kirsch wrote about John Sullivan's plot that "the key is the orange mortality rate goes from around 2,500 in August to over 5,000 in December".
However in 2018-2022, the CMR of the 80-89 age group in New Zealand ranged from about 6,635 to 7,419 deaths per 100,000 person-years, but the upper end of the age group is overrepresented in the NZ data compared to the lower end: https://mongol-fi.github.io/moar.html#Representation_of_age_groups_in_the_dataset. Based on the age composition of the vaccinated people, I got an average baseline of about 8,463 for the CMR in December 2021, which is higher than the historical CMR for ages 80-89 because ages 85-89 are overrepresented in the data compared to ages 80-84, and because I followed people who were 80-89 years old at the time of vaccination, so many of them were more than half a year older by December 2021.
When I took the first dose of each person who was 80-89 years old at the time of their dose and I followed the crude mortality rate of the cohort on each day until the end of the data, I got the same brief spike in mortality in June 2021 as Sullivan, but after that it took until June 2022 before the 14-day moving average of the CMR reached above the baseline. And even then the CMR only stayed above the baseline briefly, and June 2022 was in the middle of a COVID wave, and it was winter but I didn't account for seasonality when I calculated my baseline: https://mongol-fi.github.io/moar.html#Plot_by_John_Sullivan_for_crude_mortality_rate_by_date_in_ages_80_89.
The reason why the CMR of ages 80-89 was initially so far below the baseline could be because the healthy vaccinee effect seems to be stronger in older age groups than younger age groups.
You wrote that "The positive slope in 2023 could be due to the diminishing size of the cohort". You're probably right, because the baseline I calculated based on the age composition of the cohort also increased at a fairly steep rate in 2023, but another factor which produces the positive slope is that it was summer in early 2023 but winter in mid-2023.
In your analysis where you removed deaths in the July, August, and September of each year, you forgot to remove the person-days of alive people in summer months, which produced the two dents in mortality on either side of your "massive signal". When I reproduced your analysis, the mortality rate at the peak of the massive signal was still below the baseline based on the age composition, and when I also removed the person-days during summer months, I got rid of the two dents: https://i.ibb.co/my3zj81/massive-signal-dents.png.
Thanks Mongol.
I've quickly read through your very detailed analysis. Thanks for taking the time to document all that. I will try and reproduce UncleJohnReturns Simpson's paradox, that is indeed very interesting, (but it doesn't take away from the fact that many "fully" vaccinated elderly died during the Omicron wave in the summer of 2022 in New Zealand AND Australia. See https://kalev.substack.com/p/new-zealands-excess-deaths-a-case for New Zealand. And https://kalev.substack.com/p/australias-excess-deaths-a-case-study for Australia. Australia is particularly bad.)
> you forgot to remove the person-days of alive people in summer months
I'm not sure I follow. I completely removed all the people who died in July to September from the dataset and rebuilt the buckets from scratch and then looked at the MR. Since these people were removed from the dataset entirely they are not contributing to the alive person-day aggregates in summer months.
Yeah but the people who are alive in July-September are still added to the denominator of the mortality rate, which produces the dips in mortality rate during the weeks where winter months are overrepresented, like first around weeks 30 and 50 and a year later around weeks 80 and 100.
To avoid the bias, you could've just used the regular output of buckets.py but you could've removed all lines for July to September months, like I did in my second plot at IBB.
> To avoid the bias, you could've just used the regular output of buckets.py but you could've removed all lines for July to September months
That would introduce a bias. If you remove the July to September buckets after including everyone, the people who died in July to September would have contributed to the "alive" tally up until their death.
You can't count them as "alive" up until their death and then ignore their death.
Instead, what I did is completely remove these people from the analysis altogether, as if they never existed. As if they were never part of the dataset to begin with. They didn't contribute to any "alive" or "death" tallies.
(We are already dealing with an incomplete dataset. My understanding of Steve's approach is that an incomplete dataset does not matter. So if we remove even more records, it shouldn't matter. I did the same analysis for other age groups but it was only in the elderly where you saw an increase in MR after week 40.)
Let's not forget that the crux of the matter, and the whole reason for me writing this article, is that the total lack of transparency from many governments around the world is the real problem.
We shouldn't be left with incomplete datasets to try and draw conclusions.
As I mentioned in the article, if the evidence was overwhelmingly supportive of "vaccine safety and efficacy", they would relish demonstrating that to us. The fact that they aren't speaks volumes.
Agreed, transparency from government officials is a must.
The last few years have seen a huge increase in the lies, obfuscation, and misinformation by our “public servants” who are supposedly working for us but are if fact working for themselves and their leftist political agendas.
Great work, and many thanks for sanity checking my results.
Thank you for aggregating everyone’s analysis of the New Zealand data. This is very helpful. We are in your debt.
Great job. I totally agree. The lack of transparency tells you everything you need to know. I just emailed you give me a call and I’ll show you some things that nobody has seen yet.