I would like to start by saying that I have looked across several sites on the StackExchange website, and have determined this would be the best to ask my question as it regards data-visualisation techniques.

I recall attending a conference several years ago. The conveners were discussing methods of data visualisation and they showed a 2-dimensional scatter graph with each point being a country. And on this graph there was a reason why there were no units on the XY-axes.

Each point encapsulated several metrics of a country such as literacy rate, GDP per capita, HDI, mortality rate, income inequality etc and hence these metrics couldn't be condensed on a 2-D axis.

Yet, their was meaning behind these scattered points. Each point's distance to another point showed how similar they were to each other. I cannot recall the underyling method of measuring 'similarity' but I'm suppose correlation or some kind of clustering method was used.

I have attached a crude example drawn to show what I mean by this.enter image description here

This is purely drawn crudely, but I hope this conveys the technique used and the example case as well. Countries who's Economies are similar, such as G8 nations are closer together, and those developing such as China, Brazil and India are also closer together. Yet note the visual distance between SSA countries and G8 ones. They are far apart because they are highly distinct from each other, yet closer to developing nations such as China etc.

To summarise, the distance from any given point to another given point has meaning and is a measure of their similarity. I found this kind of graphing technique really insightful and interesting, but I just cannot remember what it is called and how to implement it for other datasets and I was wondering if anyone knows.

I hope that I have explained myself clearly, I understand that this isn't as heavy of a statistics question but I feel that this is relevant to the topics defined as suitable under this site.

  • 1
    $\begingroup$ Welcome to the site. This is a fine question for CV. Aspects of this are reminiscent of [t-sne], is that what you have in mind? $\endgroup$ Jan 7, 2021 at 17:09
  • $\begingroup$ Yes! From a brief google search, this does seem somewhat what I was looking for! Another reason I was asking was because I was wondering if such a graph could be applied to this dataset here youtube.com/watch?v=tEczkhfLwqM. An interesting animation showing how often members of congress have voted together. I was thinking instead of size of points and thickness of the edges, that the distance between any two points show how often members of congress vote together. Would t-sne be applicable to this case perhaps? $\endgroup$ Jan 7, 2021 at 17:19
  • $\begingroup$ This could be made by multidimensional scaling (I added that tag.) Search this site. $\endgroup$ Jan 8, 2021 at 3:57


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.