clustered bar graph

[Carolina Gracitelli]

Hi All,
I need to create a graph similar to this one. So I have two groups (C and G) and 3 or 4 different variables.
Do you have any suggestion?
Thank you

[Clyde Schechter]

Your picture is not visible, so we can't see what you want to do.

[Carolina Gracitelli]

I am sorry. I am attaching to you.
I need to create a bar graph with three or four different variables (e.g. divatten2 dif2 curvecoherence2 CNDA__Mean_RT_012) and 2 different groups (MVCs and no MVCs or C and G in the example).
However, when I use this command, the order is not good: For example, I would like to have the two observations of var1 (MVC and no MVCs) next to each other...
graph bar divatten2 dif2 curvecoherence2 CNDA__Mean_RT_012 if bargraph3=="Last follow-up", over(accident, relabel(2 "MVCs" 1"No MVCs" ))

Thank you

[Clyde Schechter]

I don't want to work with your complicated variable names here. But here's an example of how to do this using the built-in auto file as an example.

 clear*
sysuse auto
  // THIS GRAPH IS WHAT YOU DON'T WANT.
graph bar mpg headroom trunk, over(foreign) name(unwanted, replace)
  // YOU NEED TO RESHAPE YOUR DATA TO SEPARATE THE CLUSTERS
preserve
foreach v of varlist mpg headroom trunk {
     separate `v', by(foreign)
}
keep *0 *1
rename *# *_#
gen seq = _n
reshape long @_0 @_1, i(seq) j(vble) string
rename _0 domestic
rename _1 foreign
// THIS GRAPH IS WHAT YOU WANT
graph bar domestic foreign, over(vble) name(wanted, replace)
restore

[Carolina Gracitelli]

Dear Schechter, Thank you so much for your help. I would be able to do that without your help. The graph looks pretty nice now. I am just trying to change the names of variables in the Y axis. Thank you!

[Nick Cox]

See also statplot from SSC by Eric Booth and myself, which is focused on this need.

 
sysuse auto
statplot mpg headroom trunk , over(foreign)

[Isobel Stanley]

Hello!

I am also trying to create a graph like the above.
I have created percentage summary variables using :

by change_anaemia, sort: egen pc_stunt= mean(100*stunted)

 clear*
sysuse auto
// THIS GRAPH IS WHAT YOU DON'T WANT.
graph bar mpg headroom trunk, over(foreign) name(unwanted, replace)
// YOU NEED TO RESHAPE YOUR DATA TO SEPARATE THE CLUSTERS
preserve
foreach v of varlist mpg headroom trunk {
separate `v', by(foreign)
}
keep *0 *1
rename *# *_#
gen seq = _n
reshape long @_0 @_1, i(seq) j(vble) string
rename _0 domestic
rename _1 foreign
// THIS GRAPH IS WHAT YOU WANT
graph bar domestic foreign, over(vble) name(wanted, replace)
restore

[/QUOTE]

When I reshape my data however I get the below error

keep *0 *1

. rename *# *_#

. gen seq = _n

. reshape long @_0 @_1, i(seq) j(vble) string
(note: j = Bi Lang c c3a c3b c3c c3d c3e c3f c3g c3h c3i c3j c3k c3l c3m c3n d e f f122a field g g1221
> 2a g4b g4d h i merge oui pc_acute pc_low pc_malaria pc_stunt precision si)

variable g4b_1 type mismatch with other @_1 variables
r(198);



Would somebody be able to tell me what I am doing wrong?

Best wishes,

Isobel

[Nick Cox]

Isobel:

Short answer is No, or Not easily. You have not given us a data example we can use to see your problem. Also, this seems to be about a problem with reshape long long before you get to trying a graph command.

But type mismatch means something is numeric and something else supposedly comparable is string.

[Isobel Stanley]

Thank you Nick for your help!

https://www.statalist.org/forums/for...ered-bar-chart

[Ryan Lucas]

Clyde Schechter:

I was reading your answer above:

// THIS GRAPH IS WHAT YOU DON'T WANT. graph bar mpg headroom trunk, over(foreign) name(unwanted, replace) // YOU NEED TO RESHAPE YOUR DATA TO SEPARATE THE CLUSTERS preserve foreach v of varlist mpg headroom trunk { separate `v', by(foreign) } keep *0 *1 rename *# *_# gen seq = _n reshape long @_0 @_1, i(seq) j(vble) string rename _0 domestic rename _1 foreign // THIS GRAPH IS WHAT YOU WANT graph bar domestic foreign, over(vble) name(wanted, replace) restore

I have a similar problem, but am far less experienced. I am analysing a survey with related questions (Variables: q1-q5). The responses are binary yes/no, and can be broken up by responders' job (variable occ - there are 5 different occupations). I also want to make a bar chart with clusters of each question broken up by occupation:

I am trying to reshape this data as you suggested, but I am a little confused:

foreach V of varlist q1-q5 {
    separate `V', by(occ) 
    }

It's the next bit that I don't quite understand. Could you please explain it?

Thank you!