Superimposing a vertical line on histogram

[Demetrio Guzzardi]

Hello everyone,

I’ve been trying to figure this out for hours, but I couldn’t find an answer on this forum or elsewhere online.

I need to plot a vertical line on a histogram. However, I’m running into two issues with the xline option:

1. If the line value falls within the histogram range, the vertical line is drawn behind the histogram, so it doesn’t show clearly.
2. If the line value lies outside the automatically chosen x-range, the line doesn’t appear at all.

Here’s a minimal example that demonstrates the problem:

  sysuse auto, clear
twoway (histogram price, frac), xline(5500 20000)

At x = 5500, the red line is hidden behind the histogram, and at x = 20000, the line doesn’t appear because it’s outside the default range.

Any help would be greatly appreciated!

[Dirk Enzmann]

Use -twoway- with fcolor and xscale options:

syuse auto
twoway (histogram price, fraction fcolor(%50)), xscale(range(5000 20000)) xline(5500 20000)

[Nick Cox]

In addition to Dirk Enzmann's excellent answer, please note a miniature review

SJ-24-3 gr0097  . . . . . . . . .  Stata tip 157: Adding extra lines to graphs
        . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  N. J. Cox
        Q3/24   SJ 24(3):546--550                                (no commands)
        provides a miniature review of how you can add extra lines
        to graphs

https://journals.sagepub.com/doi/pdf...6867X241276116

This is more elaborate than might be wished, but it shows some technique.

 
 sysuse auto, clear twoway (histogram price, frac), xline(5500 20000)  twoway histogram price, frac /// || scatteri 0 5500 0.5 5500, recast(line) pstyle(p2) /// || scatteri 0  20000 0.5 20000, recast(line) pstyle(p2) /// legend(off) xtitle("`: var label price' (USD)") ytitle(Fraction)

[Demetrio Guzzardi]

Thank you both!

The issue of the overlay seem to be solved with Nick option which is a bit caothic but it works. The workaround of making the bar somewhat transparent is not quite what I need.
However both solutions imply that I manually set the ranges of the y axis. Unfortuntely I have dozens of graphs to make witithin a loop so I would prefer an option with an automatic adjustment of the range if it exist (?)

[Dirk Enzmann]

You can run sum myvar, meanonly and subquently use the r-returns r(min) and r(max) to automatically use the values needed in your twoway command, see help stored results (and the respective manual entry that is more detailed).

[Andrew Musau]

Don't you manually insert the xline positions? If so, Stata’s principle is that you can expand the range of the axis but cannot truncate it. Therefore, specifying -xscale(range(5500 20000))- should work to expand the range, and it will be ignored if the values already lie within it.

[Nick Cox]

#3 The code would look better laid out like this:

 
 sysuse auto, clear   twoway (histogram price, frac), xline(5500 20000)    twoway histogram price, frac ///  || scatteri 0 5500 0.5 5500, recast(line) pstyle(p2) ///  || scatteri 0  20000 0.5 20000, recast(line) pstyle(p2) ///  legend(off) xtitle("`: var label price' (USD)") ytitle(Fraction)

Sorry about the mess earlier.

[Nick Cox]

Arghhh! Same problem.

[Demetrio Guzzardi]

Thank you for the help. I think there was a misunderstanding, though. I would like to avoid specifying the y-axis range for the scatteri command, not the x-axis.

For the x-axis with scatteri there are no issues. The xrange option suggested by Andrew works well with xline, but the problem remains that the line extends below the histogram, while I want the line to be shown on top of the histogram.

In the end, I solved this with a workaround suggested to me here some time ago, which makes use of the undocumented twoway__histogram_gen.

I use twoway__histogram_gen to create the values of the horizontal bars in the histogram, and I then use the maximum of those values to set the y-axis in scatteri.
It’s not the cleanest solution, but it gets the job done.

Thank you everyone for the suggestions!

       sysuse auto, clear 

       twoway__histogram_gen price, frac gen(h x, replace)
       su h 
       local max_h=round(`r(max)', 0.1)
       twoway (histogram price, frac) (scatteri 0 5500 `max_h' 5500, recast(line) pstyle(p2)) (scatteri 0  20000 `max_h' 20000, recast(line) pstyle(p2))

[Andrew Musau]

For the x-axis with scatteri there are no issues. The xrange option suggested by Andrew works well with xline, but the problem remains that the line extends below the histogram

You can always get rid of the bottom margin (see highlighted).

sysuse auto, clear   
twoway histogram price, frac ///
|| scatteri 0 5500 0.5 5500, recast(line) pstyle(p2) ///  
|| scatteri 0  20000 0.5 20000, recast(line) pstyle(p2) ///  
legend(off) xtitle("`: var label price' (USD)") ///
ytitle(Fraction) xsc(r(5500 20000)) plotregion(margin(b=0))

[Andrew Musau]

Ignore #10. My interpretation of "below" was "below the value 0 on the y-axis" rather than underneath the graph. My point in #6 was that you need to specify the x-line positions manually anyway, so I didn’t see how Nick’s code introduced any additional overhead beyond that.