7 ms·

At the surface this is a fun visualization and it confirms some vague stereotypes. A closer inspection shows that the appearance of 'cultural regions' is primar

At the surface this is a fun visualization and it confirms some vague stereotypes. A closer inspection shows that the appearance of 'cultural regions' is primarily done by the use of a color scheme that is taken from beyond the data plotted. Note that India and Poland are next to each other in the data space. If you ran a clustering you would find the Indo-Polish cultural group to be very hard to split. That said, I suspect that it would be very hard to convince anyone that India is more like Poland that say Bangladesh, or conversely that Poland is more like India than say Hungary. This map is perhaps a fairly arbitrary reduction in dimensionality.

tgv 7y ago
And in the 2017 version, Poland no longer borders India, but is caught with Chile and Argentine (South Poland?), while India now is neighbors with Italy. It's certainly not a very stable mapping; perhaps all it does is amplify the noise.
- 1996 7y ago
  or, alternatively, Poland and India have seen large cultural evolutions over a short period of time.
  - tgv 7y ago
    It's not only Poland and India. And the attitudes the graph supposedly shows shouldn't change that much in 9 years.
    - 1996 7y ago
      There has been some serious changes there these last 9 years.
bhouston 7y ago
Yeah, seems very arbitrary. A multi dimensional clustering may be more useful then an attempt to flatten that with minimal distortion between the clusters would be more useful. But with too many dimensions it becomes nearly impossible....
conformist 7y ago
Agreed. Maybe colouring the dots as a function of per capita GDP would've been easy, nice, and more interesting
ianbooker 7y ago
As so often, these big survey projects lag considerably. At some point in time, these dimensions were most interesting, and it must have been very satisfying to see them explain such a big amount of variance in such a low dimensional space. I recommend the time series video: https://www.youtube.com/watch?v=ABWYOcru7js https://www.youtube.com/watch?v=ABWYOcru7js Indeed it seems less a good explanation nowadays, but this is another take away, I guess.

tgv 7y ago

And in the 2017 version, Poland no longer borders India, but is caught with Chile and Argentine (South Poland?), while India now is neighbors with Italy. It's certainly not a very stable mapping; perhaps all it does is amplify the noise.

1996 7y ago
or, alternatively, Poland and India have seen large cultural evolutions over a short period of time.
- tgv 7y ago
  It's not only Poland and India. And the attitudes the graph supposedly shows shouldn't change that much in 9 years.
  - 1996 7y ago
    There has been some serious changes there these last 9 years.

bhouston 7y ago

Yeah, seems very arbitrary. A multi dimensional clustering may be more useful then an attempt to flatten that with minimal distortion between the clusters would be more useful. But with too many dimensions it becomes nearly impossible....

conformist 7y ago

Agreed. Maybe colouring the dots as a function of per capita GDP would've been easy, nice, and more interesting

ianbooker 7y ago

As so often, these big survey projects lag considerably. At some point in time, these dimensions were most interesting, and it must have been very satisfying to see them explain such a big amount of variance in such a low dimensional space. I recommend the time series video: https://www.youtube.com/watch?v=ABWYOcru7js https://www.youtube.com/watch?v=ABWYOcru7js Indeed it seems less a good explanation nowadays, but this is another take away, I guess.