Cut axis without excluding data points

[Tarjei W. Havneraas]

Hi

I am working on a twoway plot where one data point is quite large compared to the rest. However, I do not want to exclude this point, but it would be nice to adjust the x-scale to range -.03 to .03 below, i.e. start from -.03 and not -.06.

In the graph below I have tried to illustrate what I am interested in: a graph starting at the blue line, then an arrow can indicate that the lowest data point here extends beyond -.03. It seems like -xscale(r())- does not cut the x-axis if there are data points outside the specified range.

Here is the code used to create the graph:

twoway (scatter y x, ///
xline(0) ///
xscale(r(-0.03(0.01)0.03)) ///
xlabel(-0.03(0.01)0.03) ///
(dropline x y base(0) horizontal)

[Andrew Musau]

I do not want to exclude this point, but it would be nice to adjust the x-scale to range -.03 to .03 below, i.e. start from -.03 and not -.06.

Well, this is just excluding the point. You cannot truncate the axis in twoway graphs, you need to explicitly exclude that observation using the -if- qualifier. Then you can extend the axis when plotting and include the line with an arrow (or include an invisible marker at (0 -0.03)).

[Nick Cox]

See also https://www.stata.com/support/faqs/g.../scale-breaks/

Someone commented somewhere (Twitter?) that this FAQ seems designed to make them feel guilty about wanting a scale break, which I won't fully deny. Still, there is some guidance in the FAQ.

A more exotic alternative is a cube root scale. Cube root, meaning concretely sign(x) * abs(x)^(1/3) will pull the outlier in.