Replicating Bar Graph From Article (Flors-Mas 2021)

[Alberto Burgos-Rivera]

To whomever is interested. I am trying to replicate the following bar graph I obtained from an article (Flors-Mas 2021). As you can see, this is a bar graph whose categories are distributed across different values. I was wondering how to replicated using the data used in the article which I can also attach to this post if need be. I am currently trying to replicate the aforementioned graph, but there are so many options that I don't even know where to begin.

[Nick Cox]

Please see FAQ Advice #12 about not attaching ..dta files (thanks). The version of the data here should be close enough to discuss technique.

I don't think that display is especially attractive. The use of a legend is always at best a necessary evil. There are far too many axis labels.

This is the best I could come up with using tabplot from the Stata Journal. It's reluctant to plot bars of height 0 -- or perhaps willing, but unable. If you wanted to see text of 0.0 for the zero, fudge the data value to say 1e-9. %2.1fc should get you commas as decimal points.

This is a challenge to others to produce something better.

* Example generated by -dataex-. For more info, type help dataex
clear
input float(var1 var2 percent)
1 1  1.1
2 1    0 
3 1 47.5
4 1 49.4
1 2  2.3
2 2  1.5
3 2 52.8
4 2 42.7
1 3  4.6
2 3  2.9
3 3 59.1
4 3 32.5
end
label values var1 var1
label def var1 1 "Catalan", modify
label def var1 2 `" "Catalan and" "Castilian" "', modify
label def var1 3 "Castilian", modify
label def var1 4 "Others", modify
label values var2 var2
label def var2 1 "first", modify
label def var2 2 "identification", modify
label def var2 3 "habitual", modify

tabplot var1 var2 [iw=percent], showval(format(%2.1f)) separate(var1) xtitle("") ytitle("") name(G1, replace)

[Nick Cox]

Here's another take.

* Example generated by -dataex-. For more info, type help dataex
clear
input float(var1 var2 percent)
1 1  1.1
2 1    0 
3 1 47.5
4 1 49.4
1 2  2.3
2 2  1.5
3 2 52.8
4 2 42.7
1 3  4.6
2 3  2.9
3 3 59.1
4 3 32.5
end
label values var1 var1
label def var1 1 "Catalan", modify
label def var1 2 `" "Catalan and" "Castilian" "', modify
label def var1 3 "Castilian", modify
label def var1 4 "Others", modify
label values var2 var2
label def var2 1 "first", modify
label def var2 2 "identification", modify
label def var2 3 "habitual", modify

twoway bar percent var2, barw(0.7) xla(1/3, tlc(none) valuelabel) ///
|| scatter percent var2, ms(none) mla(percent) mlabpos(12) ///
by(var1, row(2) note("") legend(off)) xtitle("") ysc(r(0 65))  name(G2, replace)

[Nick Cox]

The incomplete reference appears to expand to

Flors-Mas, A. (2021). Immigration and Languages in Catalonia: From Policies to Results. Journal of Iberian and Latin American Research, 27(2), 368–385. https://doi.org/10.1080/13260219.2021.1994712

which is behind a paywall so far as I am concerned.

https://www.statalist.org/forums/help#references explains our policy on giving full references.

[Maarten Buis]

Here is an attempt to create the original graph/table combination. It is not intended to improve on the original design.

clear
input float(var1 var2 percent)
1 1  1.1
2 1    0
3 1 47.5
4 1 49.4
1 2  2.3
2 2  1.5
3 2 52.8
4 2 42.7
1 3  4.6
2 3  2.9
3 3 59.1
4 3 32.5
end
label values var1 var1
label def var1 1 "Catalan", modify
label def var1 2 `" "Catalan and" "Castilian" "', modify
label def var1 3 "Castilian", modify
label def var1 4 "Others", modify
label values var2 var2
label def var2 1 "first", modify
label def var2 2 "identification", modify
label def var2 3 "habitual", modify

gen x = (var1 -1) * 4 + var2 -1

twoway bar percent x if var2 == 1 ||                   ///
       bar percent x if var2 == 2 ||                   ///
       bar percent x if var2 == 3,                     ///
       xlabel(1 "Catalan"                              ///
              5 `""Catalan and" "Castilian""'          ///
              9 "Castilian"                            ///
              13 "Other", notick)                      ///
       legend(off)                                     ///
       xtitle(" ", height(1.5cm))                      /// add more room below
       l1title(" ", height(1cm))                       /// add more room left
       ylabel(0(20)60)  yscale(range(0 60))            ///
       text(-13  1 "`:display %5.1f `=percent[1]''"  ) ///
       text(-16  1 "`:display %5.1f `=percent[5]''"  ) ///
       text(-19  1 "`:display %5.1f `=percent[9]''"  ) ///
       text(-13  5 "`:display %5.1f `=percent[2]''"  ) ///
       text(-16  5 "`:display %5.1f `=percent[6]''"  ) ///
       text(-19  5 "`:display %5.1f `=percent[10]''" ) ///
       text(-13  9 "`:display %5.1f `=percent[3]''"  ) ///
       text(-16  9 "`:display %5.1f `=percent[7]''"  ) ///
       text(-19  9 "`:display %5.1f `=percent[11]''" ) ///
       text(-13 13 "`:display %5.1f `=percent[4]''"  ) ///
       text(-16 13 "`:display %5.1f `=percent[8]''"  ) ///
       text(-19 13 "`:display %5.1f `=percent[12]''" ) ///
       text(-13 -2 "first"         ,placement(e))      ///
       text(-16 -2 "identification",placement(e))      ///
       text(-19 -2 "habitual"      ,placement(e))      ///
       text(-13 -2.5 "■", color(stc1))                 ///
       text(-16 -2.5 "■", color(stc2))                 ///
       text(-19 -2.5 "■", color(stc3))

[Nick Cox]

For a multi-colour version of #3

separate percent, by(var1) 

set scheme stcolor 

twoway bar percent? var2, barw(0.7 ..) xla(1/3, tlc(none) valuelabel) ///
|| scatter percent var2, ms(none) mla(percent) mlabpos(12) ///
by(var1, row(2) note("") legend(off)) ytitle("percent") xtitle("") ysc(r(0 65))  name(G2, replace)

[Nick Cox]

Yet another take (assuming data as in #2)

 
 separate percent, by(var1)  

 graph hbar percent?, over(var2) over(var1) nofill legend(off) blabel(bar) ysc(off) t1title(percent)