Coloring Histogram Bars Above/Below Thresholds

[David Kapaon]

Hi,
I'm trying to create a histogram that flags height values outside of an indicated range as red, and values within that range as green. My thought was to use the twoway function and plot two separate histograms on top of one another (one for the 'in-range' values, and one for the 'out-of-range' values). However, because the N's are different for each it throws off the bar sizes (and percentages) making the plot look odd. Is there anyway to maintain the same N for twoway plots so the scaling is consistent? I've attached some sample code below:

* Example generated by -dataex-. For more info, type help dataex
clear
input double bs041
   80
   82
   85
 87.5
   89
 89.5
 95.3
 95.5
100.7
  101
  101
  105
  105
  113
113.7
  117
120.6
120.8
120.8
122.4
  123
  125
125.9
  126
126.7
  127
  127
  130
  130
  131
  132
  132
  132
  132
  132
  132
  132
  132
  132
132.1
132.6
132.7
  133
133.7
  134
  134
  134
  134
  134
  134
  134
134.6
  135
  135
  135
  135
  135
  135
135.3
135.4
135.6
135.6
135.7
135.7
  136
  136
  136
136.9
136.9
137.5
137.7
  138
  138
138.9
139.5
  140
  140
140.2
140.6
140.7
  141
  142
  142
  142
  142
  142
  142
  142
  142
  142
  142
  142
142.1
142.3
142.7
  143
  143
  143
  143
  143
end

This is what (part) of the histogram looks like without applying different colors or restricting the observations at all

histogram bs041, percent color(grey%30) xlabel(50(50)250) ylabel(, angle(0)) xline(125 220, lcolor(black)) text(40 172.5 "Height Range (125-220 cm)")

This was my attempt - However, it makes the percent for the red bars look huge in comparison when it is just that they have different denominators

twoway histogram bs041 if (bs041 >= 125 & bs041 <= 220), percent color(green%30) || histogram bs041 if (bs041 < 125 | bs041 > 220), percent color(red%30) xlabel(50(50)250) ylabel(, angle(0)) xline(125 220, lcolor(black)) text(40 172.5 "Height Range (125-220 cm)")

I've also included a (rather crude...) picture of the larger sample as well

Thanks!
David

[FernandoRios]

Perhaps this:

ssc install frause
frause oaxaca
twoway__histogram_gen lnwage, gen(h b) w(0.1)
twoway bar h b if b<3, barw(0.1) || bar h b if b>=3, barw(0.1) || bar h b if b>=4, barw(0.1)

[Andrew Musau]

Also see https://www.statalist.org/forums/for...m-or-graph-bar

[Nick Cox]

Just a comment that combining red and green should be avoided. Difficulty in distinguishing red and green is a common visual challenge.

[ericmelse]

Following Nick's remark in #4, I recommend reading this recent paper 'open access' published in Nature Communications on the use of colour - color - in science communication, relevant for any form of visualization of quantitative data using plots, graphics: Crameri, F., Shephard, G. E., & Heron, P. J. (2020). The misuse of colour in science communication. Nature communications, 11(1), 5444. https://doi.org/10.1038/s41467-020-19160-7