I prefer median over mean in almost all situations, but you have a point there. However, these issues become less relevant after stratification by age group. I will wait with further analysis of mortality in Germany until we have reliable data for the first half of 2023.
Thanks for your comment, interesting as usual. Of course, it was not my intention to reject medians in general. After all, they serve their purpose where they are justified. I cannot share the view that age standardization would mitigate the errors, having examined the mortality risks (cases/population). They are predominantly right skewed as well. Nevertheless, I am curious about your research. Thanks also for the link. Replaced it.
Some of the median vs mean issue is due to the choice of estimator. If all you have are four (ordered) values, most estimators can be interpreted as estimators for median and mean:
Standard estimator for mean: 25%, 25%, 25%, 25%
Standard estimator for median: 0%, 50%, 50%, 0%
Harrell-Davis estimator for median: 12.66%, 37.34%, 37.34%, 12.66% (pretty close to the average of standard mean and standard median)
But what would be done with that? The key factor, in my view, is that the public expects absolute or percentage mortality figures. That places a clear requirement on the comparative value. It has to be true to the sums. Medians and even Harrett-Davis do not guarantee that.
How one divides the years (flu season, calendar year, heat weeks...) is basically irrelevant. However, since both C19 and vaccinations started around the turn of the year, calendar years are the most meaningful in this regard. The comparisons 2020 vs. previous years and 2020 vs. 2021/2022 are already sufficient to refute the whole corona and vaccination narrative, as long as you calculate halfway carefully.
I have never liked nor understood Destatis' choice of using the median value. You present an interesting exploration of the implications and I like the thought experiment.
To be honest, I have not paid attention to Destatis press releases for a long time. They usually do not deserve any attention as long as they only compare total death figures. Now that the next propaganda battle, namely the one about heat deaths due to climate change, has started in Germany, the machinations of this falsification workshop should be made public.
I am still waiting for Destatis' answer to my request on clarification of their claim that the drop in birth rates can be explained by decline in the number of women in their thirties. "Im Moment interessieren sich sehr viele Nutzende für unser Datenangebot. Wir bitten Sie deswegen um etwas Geduld und danken Ihnen sehr für Ihr Verständnis."
I had a rather different experience last time I e-mailed Destatis a couple of months ago when they called me back within a couple of days. I was quite surprised to be contacted by telephone, and even more surprised when they spent about 20 minutes talking to me and explaining the minor issue with the GENESIS online interface which I had e-mailed about (on previous occasions I had just received an e-mail response within a week)
Within the hour of me having written the above comment, I got a response (which does not mean Destatis is browsing Substack, of course...). Quite a useless response, actually, so I started the second round. I guess it makes a difference if you report technical issues or if you criticise their press statements...
I prefer median over mean in almost all situations, but you have a point there. However, these issues become less relevant after stratification by age group. I will wait with further analysis of mortality in Germany until we have reliable data for the first half of 2023.
Hagemann's video is also here, right?
https://www.bitchute.com/video/3tEF4ev47Drk/
Thanks for your comment, interesting as usual. Of course, it was not my intention to reject medians in general. After all, they serve their purpose where they are justified. I cannot share the view that age standardization would mitigate the errors, having examined the mortality risks (cases/population). They are predominantly right skewed as well. Nevertheless, I am curious about your research. Thanks also for the link. Replaced it.
Some of the median vs mean issue is due to the choice of estimator. If all you have are four (ordered) values, most estimators can be interpreted as estimators for median and mean:
Standard estimator for mean: 25%, 25%, 25%, 25%
Standard estimator for median: 0%, 50%, 50%, 0%
Harrell-Davis estimator for median: 12.66%, 37.34%, 37.34%, 12.66% (pretty close to the average of standard mean and standard median)
But what would be done with that? The key factor, in my view, is that the public expects absolute or percentage mortality figures. That places a clear requirement on the comparative value. It has to be true to the sums. Medians and even Harrett-Davis do not guarantee that.
Agree.
Hagemann's video is excellent!
- Stratify by age
- Consider demographic trends
- Do not compare months or weeks but seasons
- Work with rates, not absolute numbers
- Think about which statistics to use
I had a kind of déjà-vu at the end, when he shows his Excel sheet with ranks. :)
https://cm27874.substack.com/p/excess-deaths-in-germany-2022-update
How one divides the years (flu season, calendar year, heat weeks...) is basically irrelevant. However, since both C19 and vaccinations started around the turn of the year, calendar years are the most meaningful in this regard. The comparisons 2020 vs. previous years and 2020 vs. 2021/2022 are already sufficient to refute the whole corona and vaccination narrative, as long as you calculate halfway carefully.
Marcel Barz' video is excellent as well, and I am not only saying this because a certain Ulf Lorré has a guest appearance... :)
I have never liked nor understood Destatis' choice of using the median value. You present an interesting exploration of the implications and I like the thought experiment.
To be honest, I have not paid attention to Destatis press releases for a long time. They usually do not deserve any attention as long as they only compare total death figures. Now that the next propaganda battle, namely the one about heat deaths due to climate change, has started in Germany, the machinations of this falsification workshop should be made public.
I am still waiting for Destatis' answer to my request on clarification of their claim that the drop in birth rates can be explained by decline in the number of women in their thirties. "Im Moment interessieren sich sehr viele Nutzende für unser Datenangebot. Wir bitten Sie deswegen um etwas Geduld und danken Ihnen sehr für Ihr Verständnis."
Hi CM,
I had a rather different experience last time I e-mailed Destatis a couple of months ago when they called me back within a couple of days. I was quite surprised to be contacted by telephone, and even more surprised when they spent about 20 minutes talking to me and explaining the minor issue with the GENESIS online interface which I had e-mailed about (on previous occasions I had just received an e-mail response within a week)
Within the hour of me having written the above comment, I got a response (which does not mean Destatis is browsing Substack, of course...). Quite a useless response, actually, so I started the second round. I guess it makes a difference if you report technical issues or if you criticise their press statements...
Yes, I suspect questioning their press statements is a different matter.
And that would be really interesting if they were reading our posts!
It seems that the weather is too nice for discussion. Full reply to my 2nd enquiry: "Wir haben Ihre Einschätzung zur Kenntnis genommen."