multiple independent variables in same bar graph

[Larissa Wenger]

My data contains values of the wages of people in three different years, some of them recieved job training (program group) some didn't (control). I'd like to generate a bar graph showing the average wage of the program group next to the average wage of the control group for all three years next to each other. I couldn't find anything that works, the closest I came was using

graph bar wage1 wage2 wage3, over(prog)

but then then I have three bars of the mean of the wage of the people in the program group on the right and three bars of the mean of the wage of the people in the control group on the left. I, however, want to have mean(wage1) of program vs control to the right, then wage2 in the middle and wage 3 to the left. Is there any possibility of doing this?

Thanks a lot for any help

[Nick Cox]

Data example and graph please.

[Larissa Wenger]

I couldn't make a data example from the data I used, but this should show the same problem:

* Example generated by -dataex-. To install: ssc install dataex
clear
input int(prog engage1 engage2 engage3)
1  3  4  4
1  5  3 -5
1  6 -5 -5
1  4  4  6
0  4 -5 -5
1  4 -5  5
0  3  5  4
1  4  2  6
0 -5 -5 -5
1  5  3  6
end
label values prog PROGRAM
label def PROGRAM 0 "(0) Comparison", modify
label def PROGRAM 1 "(1) Program", modify
label values engage1 B1V3CENG
label def B1V3CENG -5 "-5/.E=missing item", modify
label values engage2 B2V3CENG
label def B2V3CENG -5 "-5/.E=missing item", modify
label values engage3 B3V3CENG
label def B3V3CENG -5 "-5/.E=missing item", modify

The first column shows the assignment program/control, the other three show variables of engagement, all collected at different points in time.

If I use

graph bar engage1 engage2 engage3, over(prog)

I get following graph:

The first problem is I'd like to have the blue bars next to each other, then the red ones and then the green ones.

The second problem I just noticed is that the -5 (indicating missing values) shouldn't be counted when the values of the graphs are calculated. I don't want to delete all observations if any of the three variables has a missing value because I could still use the ones that do exist (eg. in engage1 is missing for a particular observation it might still have values for engage2 and engage3)

Thanks for your help and sorry I didn't include everything straight away

[Nick Cox]

Thanks for the examples.

See e.g. statplot (SSC) (search the forum for mentions) and the following code

mvdecode engage?, mv(-5)

ssc inst statplot
.
statplot engage? , over(prog)
statplot engage? , over(prog) recast(bar)
statplot engage? , over(prog) recast(dot)
statplot engage? , over(prog) recast(dot) vertical
statplot engage? , over(prog) recast(dot) vertical asyvars

[Larissa Wenger]

Thank you so much