Colour Block Your Histogram - Stata Quick Tip

This short tip will show you how to colour the bars of your histogram according to value ranges along the x-axis. This allows you to colour-code different regions of your x variable. This tip was originally posted on twitter by John Mullahy (@JohnMullahy). You can find the original tweet here. It uses a Stata subroutine of the histogram command to generate the y and x data needed to create the graph.

How to Use:

note: there are two underscores between twoway and histogram below

Make a note of the bar width of your histogram, then

twoway__histogram_gen varname, generate(yvarname xvarname, replace)

twoway (bar yvarname xvarname if xvarname < first_section_top_number, color(colourname) barwidth(chosen_bar_width)) (bar yvarname xvarname if xvarname >= second_section_bottom_number & xvarname < second_section_top_number, color(colourname) barwidth(chosen_bar_width)) … … (bar yvarname xvarname if xvarname >= last_section_bottom_number, color(colourname) barwidth(chosen_bar_width))

Worked Example:

I will use the Stata auto example dataset to make a histogram of the variable length. I want to use the default Stata bar width, and there are two main ways for me to calculate this. First I can simply run a basic histogram with the command histogram length, and the pre-calculated width will be printed on the results screen. Alternatively I can calculate it by running summarize length, then use the minimum and maximum values to calculate the bar width for the standard 8 bars ((max – min)/8).

For this example I am going to use some local macros to get my bar width by the calculation outlined above. I will then separate my histogram into 3 separate bar sections. In the command pane I type the following:

sysuse auto, clear
summarize length
local wx = (`r(max)'-`r(min)')/8
twoway__histogram_gen length, generate(density len_bar, replace)
twoway (bar density len_bar if len_bar < 177, color(orange) barwidth(`wx’)) (bar density len_bar if len_bar >= 177 & len_bar < 210, color(midblue) barwidth(`wx’)) (bar density len_bar if len_bar >= 210, color(magenta) barwidth(`wx’))
drop density len_bar

This gives the following graph:

Compare this to the original look of the histogram when run by the histogram command, below: