Ordering bar graphs when using over() and by()

[Aqib Yousuf]

Hello,

I am trying to sort gender across different categories but I want the data sorted based on prespecified order. Currently, STATA is ordering the graphs based on the values on area variable(From 1 to 5). I want to order them in a way that if the gender ratio is similar then it is next to each other, and then increases based on the higher number of males or females in the next area.

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(sex area)
1 2
1 4
1 3
1 1
1 2
1 4
1 1
1 5
1 2
0 4
end
label values sex sex
label def sex 0 "Female", modify
label def sex 1 "Male", modify

------------------ copy up to and including the previous line ------------------


I am currently running the following code:

graph bar (percent), over(sex, sort(#)) by(area) asyvars

I also want to inquire if I could write a code to have all the graphs in a single row. I know that it can be done in the graph editor but having a code for it would be quite helpful as well. If you could also guide as to how I could show the number of observations using blabel as well as how I could show the total observations of each area using coding. Thank you

[Aqib Yousuf]

I also want to inquire if I could write a code to have all the graphs in a single row. I know that it can be done in the graph editor but having a code for it would be quite helpful as well. Thank you!

[Nick Cox]

The main problem here is addressed by the command myaxis from the Stata Journal which is written up at https://journals.sagepub.com/doi/pdf...6867X211045582 and also earlier at https://www.statalist.org/forums/for...e-or-graph-use

Beyond using myaxis to get the order you want, I recommend

1. In your real data you are probably using area names not numbers, and there may be many more areas than 5. In that circumstance a vertical bar chart (often called a column chart) is often a poor choice. So use a horizontal display.

2. Choosing one variable to plot, say the percent of males or the percent of females.

2. If and only if the percents in your full dataset cover a narrow range, say near 50%, a bar chart is likely to be a poor way to show the data, as comparisons with zero are not at all the issue, but comparisons of the areas with each other are the issue.

Note that as in the syntax for G2 you can use a mean to get proportions but show percents in axis labels.

Colour choices are naturally up to you, especially of colours that will make sense to (and not offend) your readership.

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(sex area)
1 2
1 4
1 3
1 1
1 2
1 4
1 1
1 5
1 2
0 4
end
label values sex sex
label def sex 0 "Female", modify
label def sex 1 "Male", modify

myaxis AREA = area, sort(mean sex)

set scheme s1color

graph bar (percent), over(sex) by(AREA, row(1)) asyvars bar(1, color(orange_red)) bar(2, color(blue)) name(G1, replace)

graph hbar sex, over(AREA) yla(0 .25 "25" .5 "50" .75 "75" 1 "100") ytitle(% of males) ysc(alt) l1title(Area) bar(1, color(blue)) name(G2, replace)

graph dot sex, over(AREA) ytitle(% of males) ysc(alt) l1title(Area) exclude0 marker(1, mcolor(blue)) name(G3, replace)

.

[Aqib Yousuf]

The main problem here is addressed by the command myaxis from the Stata Journal which is written up at https://journals.sagepub.com/doi/pdf...6867X211045582 and also earlier at https://www.statalist.org/forums/for...e-or-graph-use

Beyond using myaxis to get the order you want, I recommend

1. In your real data you are probably using area names not numbers, and there may be many more areas than 5. In that circumstance a vertical bar chart (often called a column chart) is often a poor choice. So use a horizontal display.

2. Choosing one variable to plot, say the percent of males or the percent of females.

2. If and only if the percents in your full dataset cover a narrow range, say near 50%, a bar chart is likely to be a poor way to show the data, as comparisons with zero are not at all the issue, but comparisons of the areas with each other are the issue.

Note that as in the syntax for G2 you can use a mean to get proportions but show percents in axis labels.

Colour choices are naturally up to you, especially of colours that will make sense to (and not offend) your readership.

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(sex area)
1 2
1 4
1 3
1 1
1 2
1 4
1 1
1 5
1 2
0 4
end
label values sex sex
label def sex 0 "Female", modify
label def sex 1 "Male", modify

myaxis AREA = area, sort(mean sex)

set scheme s1color

graph bar (percent), over(sex) by(AREA, row(1)) asyvars bar(1, color(orange_red)) bar(2, color(blue)) name(G1, replace)

graph hbar sex, over(AREA) yla(0 .25 "25" .5 "50" .75 "75" 1 "100") ytitle(% of males) ysc(alt) l1title(Area) bar(1, color(blue)) name(G2, replace)

graph dot sex, over(AREA) ytitle(% of males) ysc(alt) l1title(Area) exclude0 marker(1, mcolor(blue)) name(G3, replace)

. [ATTACH=CONFIG]n1702116[/ATTACH]
[ATTACH=CONFIG]n1702117[/ATTACH]

Thank you! This sorts out my issue. What other visualizations would you recommend for across area graphs?

[Nick Cox]

For what you’ve posted bar charts are widely used but in my view dot charts often work better, as already stated.

[Aqib Yousuf]

For what you’ve posted bar charts are widely used but in my view dot charts often work better, as already stated.

Is there a way to show number of observations in the bar charts? Thanks

[Nick Cox]

You started a new thread on #6 over at https://www.statalist.org/forums/for...in-a-bar-chart where I have replied too.