Graph using Stata

[Thanh Cong Gian]

Hi,

Is it possible to have this graph produced in Stata? The graph is pretty self-explanatory so I do not upload the data. Would appreciate if you can give me some hints. Thanks.

[Nathan E. Fosse]

Hi, Thanh Cong Gian -
Is it possible, yes? Is it simple to code? Nope. I haven't seen a graph as this, but I can see some far easier graphs that you could provide.

I'd do the following to make it a Stata graph:

Omit the chartjunk:

(1) the y-axis is unchanged, that could be removed, and the order of the countries could be keep from low to high values of the y-axis. There is nothing interesting in the y axis.
(2) the "blue lagoon" in the graph seems ominous but has no evidentiary value. Omit.

Graph the variation of interest (i.e., inter-country temoporal variation in the three values)

(1) dot plot of three variables by categorical variable indexing country (see Stata website for example): This could be graphed as a dot plot with countries on the y-axis and each of the 3 data points per country plotted.
(2) Multiple overlaid connected line graphs (see Stata example): line plots for each country with 3 time points for each x axis (with j country number of lines; a bit messy).

In brief, it's not going to be easy to replicate in Stata; I haven't seen graphs such as this. Also the lumpy tumor must be excised. Stata is great at many things, superior to many stats software for graphics, but to replicate this track down the R package or Python library to construct this specifically. Or make things simple.

Cheers,
- Nate

[Nick Cox]

The horizontal lines with data points will yield to twoway rspike together with scatter or scatteri.

The rest of the data, the point cloud, is a set of points on another scatter plot, or else not explained well enough.

[Nathan E. Fosse]

Hi, Thanh Cong Gian -
Is it possible, yes? Is it simple to code? Nope. I haven't seen a graph as this, but I can see some far easier graphs that you could provide.

I'd do the following to make it a Stata graph:

Omit the chartjunk:

(1) the y-axis is unchanged, that could be removed, and the order of the countries could be keep from low to high values of the y-axis. There is nothing interesting in the y axis.
(2) the "blue lagoon" in the graph seems ominous but has no evidentiary value. Omit.

Graph the variation of interest (i.e., inter-country temoporal variation in the three values)

(1) dot plot of three variables by categorical variable indexing country (see Stata website for example): This could be graphed as a dot plot with countries on the y-axis and each of the 3 data points per country plotted.
(2) Multiple overlaid connected line graphs (see Stata example): line plots for each country with 3 time points for each x axis (with j country number of lines; a bit messy).

[wbuchanan]

I'd also agree that with what others have previously stated. One thing that could help when developing your graphs is to think about what data each visual element is encoding. So, the "blue lagoon" as Nathan E. Fosse calls it, is encoding some data but the reader would not have any way to determine what precisely is being encoded by the shape, the color, and the combination of the shape and color. Another consideration is that you are encoding the same periods in time with different symbols that is easily conflated with the meaning of the data encoded on the x-axis. In several cases, the color encoding of the countries is too similar to distinguish the countries (e.g., Vietname and Philippines appear to use the same color); however, it might be more useful to use fewer color encodings (e.g., for the different geographical regions) and then using varying line patterns (e.g., solid, dots, dashes, etc...) to show the countries within those regions. It wouldn't be idael, but could get you a bit more bang for the buck with regards to being able to more easily compare across countries and regions simultaneously. It also seems a bit counter intuitive that the "growth rate of per capita GDP during 1970-2013" would be completely flat for each of these countries. In short, it seems that there is too much information trying to be encoded with too few dimensions to make it unambiguous.

[Thanh Cong Gian]

Thanks a ton, guys. I thought like Nathan- the dot plot- when I started working on this but the Stata commands from Nick is brilliant. works like a charm. I agree the blue lagoon is ambiguous and I don't know what that really means. I figured it's a x-country scatter plot between growth rate and the starting GDP per capita (decreasing trend line) but the distribution is restricted. wbuchana, I agree with you about the potential ambiguity because we want to convey so much information given the limited dimensions we have. But this chart looked pretty cool to me when I first saw it. Cheers