Graphing charlson comorbidity index by group and decade

[LarsFolkestad]

Dear listers

I have a dataset with the following variables
Id, decade, CCI and group
Every id can be CCI=0,1,2,3
I would like to plot by group and decade the fraction of id's that is 0's / 1's/2's/3's

I would like it to be stacked bars if possible, and with the sum being 100% pr bar.

Lars

[LarsFolkestad]

i ended up doing the following and that seemed to work.

gen CCI0=1 if CCI==0
gen CCI1=1 if CCI==1
gen CCI2=1 if CCI==2
gen CCI3=1 if CCI==3

collapse (count) CCI0 CCI1 CCI2 CCI3, by(group decade)
graph bar CCI0 CCI1 CCI2 CCI3, over (decade, gap(10)) percentages stack by(group)

[Nathan E. Fosse]

Without knowing more of the data, this seems like a good approach. The problem with the collapse command is that it: (1) renames the variables of interest; and, (2) replaces the data in working memory with the collapsed data. It doesn't sound as if (1) is an issue for your but you may want to use the preserve and restore commands, as below.

I'd add to your syntax as follows a few other tweaks (see below).
Good luck!
- Nate

preserve // keeps data in working memory intact

CCI // number of distinct values of CCI
loc i_max r(ndistinct) // create local equal to number of distinct values of CCI

* Generate variables using loops instead
forvalues i = 0(1)`i_max' {
gen CCI`i' = 1 if CII == `i'
}

* Collapse
collapse (count) CCI`i'-CC`i_max', by(group decade)

* Create bar graph
graph bar CCI`i'-CC`i_max', over (decade, gap(10)) percentages stack by(group)
graph export graphname, as(png) // export it

restore // returns to data in working memory
d,s // short description of data in working memory