Bar graphs with brackets to indicate statistical significance

[Tugba Atasoy]

Dear statalist members,

I would like to produce a bar graph with the (mean) differences for two different variables between two treatment groups. Producing this graph is not a problem; however, I would like to indicate the statistical significance of the (mean) differences between the treatments for each variable by using brackets and some text_options (?) (i.e., n.s. or the corresponding asterisks and/or the p-values). I attached two exemplary bar graphs, which I found online from other forums.

How can I modify the following code to insert such brackets?
graph bar variable_a variable_b, over(treatment) ///
legend(label(1 "aa") label(2 "bb")) ///

I would greatly appreciate any help on this.

Many thanks and best,
Tugba

[Andrew Musau]

You won't be able to do this using the graph command but with some work, you can do it with twoway. See the following:

https://www.statalist.org/forums/for...raphs-in-stata

[Tugba Atasoy]

Dear Andrew,

thank you very much for your suggestion. I eventually decided to update the figure with the free vectors graphic editor Inkscape.

Best,
T

[Retselisitsoe Pokothoane]

Dear Stata Users,

I want to draw the bar chart showing the cigarette brands smoked by gender. However, I do not know how to show significant differences in cigarette brands smoked between males and females? How can I show the significant differences on a bar graph? The cigarette brand is a categorical variable of 5 brands of cigarettes, and gender is a dummy variable which equals 1 if a respondent is male and 0 if a female.

Regards,
Pokothoane

[Nicholas Owsley]

I would really like to follow-up on Tugba's post? I appreciate Andrew's redirect, but I'm just not sure how to do the same thing on the bar twoway space and it also seems like a really painful workaround.

Could anyone help? I'd like to make exactly the same graphs Tugba wanted.

[Andrew Musau]

Two scatteri calls, one with option -recast(line)- and the second with option -recast(dropline)- and a defined base should do it.

sysuse pop2000, clear
replace maletotal = maletotal/1e+6
tw (bar maletotal agegrp in 1/5) ///
   (scatteri 10.75 2 10.75 3,  recast(line) lw(medthick)  mc(none) lc(black) lp(solid)) ///
   (scatteri 10.75 2 10.75 3,  recast(dropline) base(10.6) lw(medthick) mc(none) lc(black) lp(solid)), ///
   ysc(r(., 11)) scheme(s1mono) leg(off)

[Nicholas Owsley]

Thank you, Andrew!

[smith Jason]

Where are the numbers "
scatteri 10.75 2 10.75 3, " from? People want to learn and never understand it.

[Nick Cox]

scatteri 10.75 2 10.75 3, recast(line)

comes from looking at the graph and thinking that a straight line is needed between (10.75, 2) and (10.75, 3).

In #6 find the horizontal y = 10.75 and the vertical x = 2 and where they intersect as the point (10.75, 2) -- and so on.

A scatteri call without a recast() would give you individual points on a scatter plot. recast(line) converts two points to a line segment.

They are user choices, not results of other commands.

[smith Jason]

This code add bracket to the graph, however, the significance is still not available.

[Nick Cox]

Indeed. You need extra code to write text on the graph. See

help added text options

[smith Jason]

Thank you! I really tried with the code below, however, the number didn't come out.
clear
sysuse pop2000, clear
replace maletotal = maletotal/1e+6
tw (bar maletotal agegrp in 1/5) (scatteri 10.55 2 10.55 3, recast(line) ///
lw(medthick) mc(none) lc(black) lp(solid)) (scatteri 10.55 2 10.55 3, ///
recast(dropline) base(10.52) lw(medthick) mc(none) lc(black) lp(solid)), ///
ysc(r(., 11)) scheme(s1mono) leg(off) ///
text(2.5 10.52 "6789", place(n))

[Nick Cox]

You have the coordinates the wrong way round.

[smith Jason]

I am sorry that I really don't know how to add the texts above the flat line although I adjusted the coordinate to make the graph beautiful.

[Nick Cox]

The text should be at 10.52 2.5 presumably not 2.5 10.52.