How to ask stata to round the labels in a histogram?

[Stefano Bovio]

Hi ,
I am using the following code to make a graph, but: the decimals in the graph are annoying; if I sum the numbers the result does not give 100%.
Is there a way to round the decimals in the label?

histogram NormSocialCapitalINDEX if SoluzioneAbitativa==3, width(10) start(0) percent addlabel normal ylabel(0(2)38) xlabel(0(10)100) color(green) lcolor(black) addlabopts(mlabsize(medsmall))

Thanks

[Ken Chui]

Use option yvarformat, like this:

sysuse auto, clear

histogram mpg if foreign == 1, width(10) start(0) percent addlabel normal ylabel(0(2)38) xlabel(0(10)100) color(green) lcolor(black) addlabopts(mlabsize(medsmall) yvarformat(%4.1f))

[William Lisowski]

To what Ken wrote I will add that you can never be sure that the individual percentages displayed will add to 100, because of rounding. Consider 99 observations with 33 values of 1, 33 values of 2, and 33 values of 3. The percentages will all be displayed as 33 or 33.3 or ... and will sum to 99 or 99.9 or ... .

[Nick Cox]

I had cause to cite https://statweb.stanford.edu/~cgates...freedman79.pdf in a recent thread, and here we are again.

[Stefano Bovio]

Thank you Ken, it work perfectly.
But William: is it normal that the sum does not make 100 even if I do not round? This is the thing that weird me

And thank you Nick for the advice...

[Stefano Bovio]

And a curiosity: where is it that I find the code to round to two decimal or the other rounding possibilities?

[Nick Cox]

Use a different display format to round to a different number of decimal places.

[William Lisowski]

is it normal that the sum does not make 100 even if I do not round? This is the thing that weird me

Yes it is normal. It is a matter of precision. Just like you can only write 100/3 to a limited number of digits when it is actually 33.3333... continuing forever, Stata can only store a limited number of digits. And when you use pen and paper to write 33.333...3 three times and then add them together, you will get 99.999...9 as a result.

For a good discussion of precision in Stata and the issues involved see the output of

help precision

and this discussion on the Stata blog.

https://blog.stata.com/2012/04/02/th...-to-precision/