It’s not exhaustive but you can see average speeds for people on starlink on their website. Starlink speeds vs fiber speeds will be driven by different technological hurdles but a pretty good data point for this kind of analysis.
Question on much of the data here is "How much political bias is factored into the data?" Everything Google touches gets tainted so how much can you trust the data.
FYI: An interesting graph would be internet speed versus political slant of the area. Maybe just a US graph of speed in congressional districts versus party controlling the district.
For one, you should never trust a source that can be edited by just anyone, and not accepted as a source by any US Court or any US Learning institution. Secondly, only a fraction of America even accepts "green movement" theology (yes, it is a religion) as fact. Lastly, yes, the more people online at any given time will slow internet speeds, that is a known fact. You want faster speeds, pay for it. Or in Socialist countries, they need to tax the population more to invest more in their infrastructure so they can put more data lines in or higher speed lines in.
Emil, this is off topic, but it would be great to hear some commentary from you on the new David Reich paper and also the most recent review of admixture studies:
The two columns in the table you show are wildly incompatible with each other. Looking at the figures for Denmark, the average download speed is given as 49.98 mbit/s and the median download speed is given as 224.14 mbit/s.
The problem is that this is completely impossible. It's not even close to being explainable by an error somewhere. Suppose there are 2X internet connections in Denmark. We'll round the average download speed up to 50, and heck, let's round the median download speed down to 150.
Now, since the average download speed is 50 mbit/s, if every connection in Denmark spends one second downloading, they will download a total of 2X*50 = 100X megabits of data.
And since the median download speed is 150... you know what, let's assume that in addition to being the median, 150 is also the maximum. So the X internet connections above the median speed have an average speed of 150 mbit/s, meaning that if 𝘁𝗵𝗲𝘆 each spend one second downloading, they'll collectively download a total of 1X*150 = 150X megabits of data.
Except of course that this exceeds the amount that would be downloaded if the other X connections, the ones below the median download speed, also spent one second downloading. So we've learned that the average download speed among those connections below the median is quite sharply negative. You learn something new every day.
Urbanization rate data isn't very good, because the individual countries determine what is considered "urban". The definition of urban varies pretty widely by country. This leads to stuff like Egypt having a much lower urbanization rate than the US, even though Egyptians live much more concentrated than Americans.
As you've pointed out yourself, weighted density is by far the best metric.
It’s not exhaustive but you can see average speeds for people on starlink on their website. Starlink speeds vs fiber speeds will be driven by different technological hurdles but a pretty good data point for this kind of analysis.
https://www.starlink.com/map?view=download
Emil--you have too much time on your hands. :) love your stuff.
Looking at my files, this project took 3 hours from downloading the data, writing 450 lines of R code, and writing the post.
Then, you have too much time on your hands and are also highly efficient. That is a frightening combination.
Nice article, thanks.
Question on much of the data here is "How much political bias is factored into the data?" Everything Google touches gets tainted so how much can you trust the data.
FYI: An interesting graph would be internet speed versus political slant of the area. Maybe just a US graph of speed in congressional districts versus party controlling the district.
For one, you should never trust a source that can be edited by just anyone, and not accepted as a source by any US Court or any US Learning institution. Secondly, only a fraction of America even accepts "green movement" theology (yes, it is a religion) as fact. Lastly, yes, the more people online at any given time will slow internet speeds, that is a known fact. You want faster speeds, pay for it. Or in Socialist countries, they need to tax the population more to invest more in their infrastructure so they can put more data lines in or higher speed lines in.
Emil, this is off topic, but it would be great to hear some commentary from you on the new David Reich paper and also the most recent review of admixture studies:
https://www.qeios.com/read/5OULH0
The two columns in the table you show are wildly incompatible with each other. Looking at the figures for Denmark, the average download speed is given as 49.98 mbit/s and the median download speed is given as 224.14 mbit/s.
The problem is that this is completely impossible. It's not even close to being explainable by an error somewhere. Suppose there are 2X internet connections in Denmark. We'll round the average download speed up to 50, and heck, let's round the median download speed down to 150.
Now, since the average download speed is 50 mbit/s, if every connection in Denmark spends one second downloading, they will download a total of 2X*50 = 100X megabits of data.
And since the median download speed is 150... you know what, let's assume that in addition to being the median, 150 is also the maximum. So the X internet connections above the median speed have an average speed of 150 mbit/s, meaning that if 𝘁𝗵𝗲𝘆 each spend one second downloading, they'll collectively download a total of 1X*150 = 150X megabits of data.
Except of course that this exceeds the amount that would be downloaded if the other X connections, the ones below the median download speed, also spent one second downloading. So we've learned that the average download speed among those connections below the median is quite sharply negative. You learn something new every day.
Urbanization rate data isn't very good, because the individual countries determine what is considered "urban". The definition of urban varies pretty widely by country. This leads to stuff like Egypt having a much lower urbanization rate than the US, even though Egyptians live much more concentrated than Americans.
As you've pointed out yourself, weighted density is by far the best metric.