How to tweak a bar graph

[Claude Francoeur]

Hello,

I am using the following commands to produce the graph below:

#delimit ;
graph bar, over(COUNTRY) over (PEMP1921cat)
asyvars
blabel(bar, format(%9.1f))
percentages
bar(1, color(blue))
bar(2, color(orange))
ytitle("Number of companies");

For each of the three categories of PEMP1921cat, the bars show the percentage of companies in each country (CAN or USA) relative to the total number of companies in both countries, for a total of 100%.

I would prefer having either the number of companies for each country by category or having the percentage of companies in each category relative to the total number of companies in total for that same category. In other words, if there were 100 Canadian companies in total that were distributed 20, 40, 40 in the three categories, I would like the bars to show either the counts (20, 40, 40) or, even better, the corresponding percentages, i.e. 20%, 40% and 40%. Same for the US companies.

I tried different commands without success. Is this feasible?

Thank you!

[Jared Greathouse]

Data example please.

[Hemanshu Kumar]

And you aren't happy with the results of just switching the order of your -over- options?

Compare

sysuse auto, clear
#delimit ;
graph bar, over(rep78) over(foreign) 
    asyvars 
    blabel(bar, format(%9.1f)) 
    percentages
    scheme(s2color)
    ;
#delimit cr

which produces:

with the results from

sysuse auto, clear

#delimit ;
graph bar,  over(foreign) over(rep78) 
    asyvars 
    blabel(bar, format(%9.1f)) 
    percentages
    scheme(s2color)
    ;
#delimit cr

which produces:

[Claude Francoeur]

No, switching the options does not produce what I want. Please find the data attached. Thanks!

[Jared Greathouse]

No no no, I mean a data example using dataex

[Hemanshu Kumar]

No, switching the options does not produce what I want. Please find the data attached. Thanks!

Well, then I'm not quite sure what you want. Switching the options does produce a graph with three bars each for CAN and USA, with their respective distributions over the three categories of PEMP1921cat. If that's not what you want, you'll need to explain what you want different.

[Nick Cox]

Jared Greathouse is alluding to the longstanding request at https://www.statalist.org/forums/help#stata to post examples using dataex.
You're asked to read the FAQ Advice before posting.


For graphical purposes, Claude's dataset with 3713 observations is equivalent to this result from contract

* Example generated by -dataex-. For more info, type help dataex
clear
input str3 LOC float PEMP1921cat int _freq
"CAN" 0  181
"CAN" 1  140
"CAN" 2  153
"USA" 0 1056
"USA" 1 1098
"USA" 2 1085
end
label values PEMP1921cat PEMP1921cat
label def PEMP1921cat 0 "-.6473907-", modify
label def PEMP1921cat 1 "-.0597015-", modify
label def PEMP1921cat 2 ".1069561-", modify

Then here is some code. I used catplot from SSC.

catplot LOC PEMP1921cat [fw=_freq] , recast(bar) percent(LOC) asyvars bar(1, lcolor(blue) fcolor(blue*0.5)) bar(2, lcolor(red) fcolor(red*0.5)) blabel(total, format(%2.1f)) t1title(explain the categories somewhere)

[Claude Francoeur]

That works fine. Thank you Nick! And I will use dataex next time.