Bar graph comparing full sample mean with means by group

[Youssef Chouhoud]

Hi everyone,

I'm not quite sure how to phrase this request, but I want to create a bar graph that shows the mean of a full sample and compare it to the mean of groups within the sample in *one* graph. In the attached image, you see the far right bar represents the general population and the remaining graphs represent the major religious groups in the gen pop. I'd like to recreate something along these lines. Many thanks for the help.

[eric_a_booth]

A toy example to get you started:

sysuse auto, clear

*create total line:
collapse (mean) price
g rep78 = 99
sa total, replace
sysuse auto, clear
collapse (mean) price, by(rep78)
append using total

lab def rep78 1 "test" 2 "test" 99 "Overall", modify

lab val rep78 rep78
list
gr bar price, over(rep78) 

*set specific bar colors

graph bar (mean) price, over(rep78) ///
 ascategory asyvars showyvars ///
bar(1, fcolor(gs1)) ///
bar(2, fcolor(gs1)) ///
bar(3, fcolor(gs1)) ///
bar(4, fcolor(gs1)) ///
bar(5, fcolor(gs1)) ///
bar(6, fcolor(red)) ///
bargap(20) blabel(bar)

Aside: Is the general public bar really the same as the 'total respondents' base bar described in the note?

[Youssef Chouhoud]

Thanks a bunch, Eric! This is definitely enough to get me going where I want to end up.

As for your question, I think you astutely noticed that clearly there are folks in America that do not fit neatly into the graph's categories. The analysis for that visualization omitted the "other" and "dk/ref" cats. Was that what you were getting at?