Ciplot for multiple groupvars

[Mattia Gatti]

Dear All,

I am trying to create a series of ciplots displaying the mean value and ci for different groups. Unfortunately it appears ciplot allows just for one groupvar with the by() command.

Here is my dataset:

* Example generated by -dataex-. For more info, type help dataex
clear
input long country float idno str15 party float(vote dist_lr dist_red dist_immpoor) byte gender float age
 3  3 "AfD"    0 4.2380953         6       9.9 0 2
 3  3 "SPD"    0 1.3809524 3.0526316      4.15 0 2
 3  3 "CDU"    1  .8571429  5.947369         6 0 2
 3  3 "LINKE"  0 3.5714285  .8947368       2.7 0 2
 3  3 "FDP"    0 1.4285712  8.578947      5.95 0 2
 3  3 "GRUNEN" 0 1.7619047  3.421053       1.7 0 2
10  6 "GP"     0  1.857143        .5  3.333333 1 1
10  6 "FF"     0  .4285712  2.357143  1.833333 1 1
10  6 "Lab"    0 1.4285715  .7857144  2.833333 1 1
10  6 "SF"     0 2.4285715  .2142856  2.333333 1 1
10  6 "FG"     1  .8571429  3.357143         2 1 1
 6  7 "SFP"    0         0 2.2142859 1.0476189 1 3
 6  7 "SDP"    0  2.857143  .7857143  .3809524 1 3
 6  7 "VIHR"   0 2.2857144  .3571429 2.0476189 1 3
 6  7 "KESK"   0 .14285707  2.428571 2.2380955 1 3
 6  7 "KD"     0  .8571429 2.2857141 2.4358976 1 3
 6  7 "PS"     0 1.5714288  2.785714  6.452381 1 3
 6  7 "KOK"    1  1.714286  4.357143  1.095238 1 3
 6  7 "VAS"    0       4.5 1.4285715 1.1190476 1 3
 7  9 "PS"     1         1      2.25 .58333325 1 3
 7  9 "LR"     0     5.875     7.375  4.541667 1 3
 7  9 "LREM"   0 4.3333335  5.833333 2.3333333 1 3
 7  9 "MoDem"  0     4.125  5.857143 2.0416667 1 3
 7  9 "FI"     0       .75      .625  .6666667 1 3
 7  9 "EELV"   0        .5     2.375 1.7083333 1 3
 7  9 "RN"     0      7.75         4  6.541667 1 3
 8 10 "LIBDEM" 0 .23529434  .9705882  .7450979 1 1
 8 10 "UKIP"   0    4.6875 4.5666666  6.137256 1 1
 8 10 "LAB"    0 2.0588236  .9117647 .19607854 1 1
 8 10 "SNP"    0        .5      .625         1 1 1
 8 10 "GREEN"  0         2      .625  1.627451 1 1
 8 10 "CONS"   1  3.117647  4.147059  4.254902 1 1
12 11 "VVD"    0 2.6153846       .75 4.5897436 0 1
12 11 "D66"    1 .15384626      1.75 .33333325 0 1
12 11 "SP"     0 3.6153846      6.75 1.9166667 0 1
12 11 "PVV"    0 3.6923075 1.6666665  6.589744 0 1
12 11 "PvdA"   0 1.3846154  4.666667  .8333333 0 1
12 11 "CDA"    0 1.8461537       1.5  3.666667 0 1
12 11 "CU"     0  .0769229  3.318182 1.1282051 0 1
12 11 "GL"     0  2.692308  5.318182 1.7179487 0 1
 1 13 "SPO"    0         5       2.4 1.3666666 1 1
 1 13 "OVP"    0       2.1       6.2  5.266667 1 1
 1 13 "Grune"  0       6.5       2.5  .6333332 1 1
 1 13 "FPO"    1 .10000038       5.6  6.466667 1 1
 1 13 "NEOS"   0       3.1       6.9  .9666669 1 1
11 14 "FdI"    0 2.0526314  3.029412 3.1754384 1 0
11 14 "LN"     0 1.7894735   4.81579  3.280702 1 0
11 14 "FI"     1 .05263138  4.388889  .3333335 1 0
11 14 "PD"     0  3.789474 .22222233  3.614035 1 0
11 14 "M5S"    0 2.2222223  .6052631 .11111116 1 0
11 14 "SI"     0  5.555555       1.5  5.979167 1 0
 7 17 "RN"     0      2.75         1 3.2083335 0 0
 7 17 "PS"     0         4      2.75 3.9166665 0 0
 7 17 "EELV"   0       4.5     2.625  5.041667 0 0
 7 17 "FI"     0      5.75     4.375 2.6666665 0 0
 7 17 "LREM"   0  .6666665  .8333335         1 0 0
 7 17 "MoDem"  0      .875  .8571429 1.2916665 0 0
 7 17 "LR"     1      .875     2.375 1.2083335 0 0
 6 21 "VIHR"   0 1.2857144  .3571429  5.380952 1 0
 6 21 "KOK"    1  2.714286  4.357143 2.2380953 1 0
 6 21 "PS"     0  2.571429  2.785714 3.1190476 1 0
 6 21 "SFP"    0         1 2.2142859  4.380952 1 0
 6 21 "VAS"    0       3.5 1.4285715  4.452381 1 0
 6 21 "SDP"    0  1.857143  .7857143  2.952381 1 0
 6 21 "KESK"   0  .8571429  2.428571 1.0952377 1 0
 6 21 "KD"     0  1.857143 2.2857141  .8974357 1 0
11 22 "FdI"    0  7.052631  3.029412  9.842105 0 0
11 22 "LN"     0  6.789474   4.81579  9.947369 0 0
11 22 "SI"     0  .5555556       1.5     .6875 0 0
11 22 "M5S"    0  2.777778  .6052631  6.555555 0 0
11 22 "PD"     1 1.2105262 .22222233 3.0526316 0 0
11 22 "FI"     0  4.947369  4.388889         7 0 0
 3 23 "CDU"    0  2.857143  5.947369         6 0 3
 3 23 "LINKE"  0 1.5714285  .8947368       2.7 0 3
 3 23 "AfD"    0  6.238095         6       9.9 0 3
 3 23 "GRUNEN" 0  .2380953  3.421053       1.7 0 3
 3 23 "SPD"    1  .6190476 3.0526316      4.15 0 3
 3 23 "FDP"    0  3.428571  8.578947      5.95 0 3
 3 24 "SPD"    1 1.6190476  .5526316  .8166668 1 2
 3 24 "GRUNEN" 0 1.2380953  .9210527 1.6333332 1 2
 3 24 "AfD"    0  7.238095       3.5  6.566667 1 2
 3 24 "FDP"    0  4.428571  6.078947 2.6166666 1 2
 3 24 "LINKE"  0 .57142854 1.6052632  .6333332 1 2
 3 24 "CDU"    0  3.857143 3.4473686  2.666667 1 2
10 25 "GP"     0  3.142857         3  3.333333 0 1
10 25 "SF"     0 2.5714285 2.2857144  2.333333 0 1
10 25 "FG"     0  5.857143  5.857143         2 0 1
10 25 "FF"     1  5.428571  4.857143  1.833333 0 1
10 25 "Lab"    0 3.5714285 3.2857144  2.833333 0 1
10 26 "FG"     0  .8571429  1.642857  5.333333 1 2
10 26 "Lab"    0 1.4285715  4.214286  6.166667 1 2
10 26 "FF"     1  .4285712  2.642857  5.166667 1 2
10 26 "GP"     0  1.857143       4.5  6.666667 1 2
10 26 "SF"     0 2.4285715  5.214286  5.666667 1 2
 5 27 "Cs"     0       7.2  6.533333  6.428571 1 2
 5 27 "Vox"    0  9.714286  8.333333       9.8 1 2
 5 27 "Pais"   0  2.733333 2.3333333 2.2857144 1 2
 5 27 "IU"     0 1.8666667       1.4  2.533333 1 2
 5 27 "PSOE"   1       3.6       2.8 3.9333334 1 2
 5 27 "PPP"    0  8.066667  7.266667       7.6 1 2
end
label values country country_label
label def country_label 1 "1. Austria", modify
label def country_label 3 "3. Germany", modify
label def country_label 5 "5. Spain", modify
label def country_label 6 "6. Finland", modify
label def country_label 7 "7. France", modify
label def country_label 8 "8. Great Britain", modify
label def country_label 10 "10. Ireland", modify
label def country_label 11 "11. Italy", modify
label def country_label 12 "12. Netherlands", modify
label values gender gender
label def gender 0 "0. Female", modify
label def gender 1 "1. Male", modify
label values age age_groups
label def age_groups 0 "0. 30 and Under", modify
label def age_groups 1 "1. 31-50", modify
label def age_groups 2 "2. 51-69", modify
label def age_groups 3 "3. 70 and over", modify

Here is the code I am using:

 ciplot dist_lr if vote==1, by(gender)

And here is what I would like to do:

 ciplot dist_lr if vote==1, by(gender age education)

I had a look at previous forum conversations and at the help window for ciplot where statsby is suggested for more flexibility. Yet, I have not figure out how to make it work for my specific case.

Lastly, I would then like to do the same but for different variables (eg. dist_red dist_immdfrace). Perhaps I could then combine the graphs?

Sincerely
Mattia

[Nick Cox]

I have tried hard to convince people that ciplot (from SSC) is no longer to be used unless it covers your use case. But if you create a composite

egen group = group(gender age education), label

then you have just one group variable. See also https://www.stata-journal.com/articl...article=dm0034

That could easily get messy, as the number of categories concerned might well be 50 to 100, but it's what you're asking for. But in turn combining that graph with yet others is predictably yet messier.

Is that dataset your entire dataset, or just an example?

[Mattia Gatti]

Dear Nick,

thanks for the advice. What should I use instead of ciplot? As for the "group" variable, this is not my intended purpose as it creates different combinations of the three specified variables. I am instead trying to get means for the three variables, separately, yet in the same plot. This means that, if for gender I have "male" and "female", for education "low", "medium", "high", and for age "under 30" "30-60" "60 and over", I would like to get 8 means (and CIs). Sorry if I was not clear before.

The dataset presented here is just a part of the entire dataset.

Sincerely
Mattia

[Nick Cox]

That's not the standard meaning of a by() option in Stata. The standard meaning is, as you found, that you get cross-combinations of two or more variables.

But wanting them by(a b c) to mean by(a) AND by(b) AND by(c) -- although by(a b c) won't do it -- makes sense and can be done.

Here's a sketch. Different dataset, but one outcome and three categorical predictors.

set scheme stcolor

foreach v in race collgrad south {
    webuse nlswork, clear
    levelsof `v', local(levels)
    su `v', meanonly
    local min = r(min) - 0.3
    local max = r(max) + 0.3
    statsby, by(`v') : ci means ln_wage

    scatter mean `v' || rcap ub lb `v', name(`v', replace) xsc(r(`min' `max')) xla(`levels', valuelabel) legend(off)
}

graph combine race collgrad south, row(1) ycommon l1title(ln_wage)

[Mattia Gatti]

Dear Nick,

It works perfectly. Thanks a lot.

Sincerely
Mattia