Dotplot grouped by third variable

[amandacpac]

Hello Statalisters! I am trying to make a dotplot of a continuous variable over values of a string variable. That part is easy enough, using the "over()" statement. However, I would like to see these all grouped by the levels of a third variable. I don't necessarily want separate plots, though I will settle for that if necessary, but rather to have a small break in the x-axis that distinguishes each group (similar in look to a coefplot after mlogit, but vertical and without the error bars like the plot below)



Using the automobile data as an example, imagine I would like something like a dotplot of mpg over make of automobile (see code below), but with all the foreign cars on the left side of the plot, with a label of "Foreign" under them, and all the domestic cars on the right side of the plot with the word "Domestic" under them. In my actual data, I have 6-10 groups and will be repeating this a few times with the groups changing, so, making this somewhat automated will be very helpful.

sysuse auto, clear
dotplot mpg, over(make) xlab(,angle(90))

[Andrew Musau]

A dot plot is just a special kind of scatter plot. With this in mind, you can use the more flexible twoway command for this. For your example, I reduce the number of categories (makes of cars) and I place the categories on the y-axis to enhance readability. Additionally, I use labmask from the Stata Journal, authored by Nick Cox.

sysuse auto, clear
drop in 20/60
dotplot mpg, over(make) xlab(,angle(90))
gen Make=cond(foreign==0, _n*5, 10+(_n*5))
labmask Make, values(make)
tw (scatter Make mpg if !foreign, ylab(5(5)95 110(5)175, val angle(horiz) ///
labsize(vsmall))) (scatter Make mpg if foreign, scheme(s2color) ///
ytitle("Domestic                                Foreign") leg(off))

Res.:

[amandacpac]

This is great, but, I have two datasets, one with 6 groups and one with 10 groups. The groupings change based on other analyses, so, I would like to be able to run with new groupings. Most of the code will work ok for that, except the labeling of "Domestic" and "Foreign". Is there a way to automate that to automatically pull in the values of that variable?

[Nick Cox]

As a variant on Andrew Musau's example consider

sysuse auto, clear
drop in 20/60

graph dot (asis) mpg, over(make, sort(1) label(labsize(small))) over(foreign) nofill exclude0 ysc(r(12, .) alt)

With graph dot I tend to add linetype(line) lines(lc(gs12) lw(vthin)) for various reasons, one being that I've found that the dotted grid doesn't transport well to other software.

[Andrew Musau]

Most of the code will work ok for that, except the labeling of "Domestic" and "Foreign". Is there a way to automate that to automatically pull in the values of that variable?

These are just axis titles and you will have to place them manually. However, see

help title_options

This will allow you to choose where to place the group labels and other options, turning off the defaults.

[amandacpac]

Nick Cox This is perfect! However, is there a way to present this horizontally? So, that the make of car runs along the x-axis and the mpg along the y-axis? I have looked through the documentation and don't see an option for it, but, maybe I'm missing something. If not, I can leave it like this. Thank you!

[Nick Cox]

graph dot has an undocumented vertical option which usually makes displays much uglier. That's my guess at why it is undocumented.