line graph of row percentage by different groups

[Sherine Maui]

Hi Statalist,

I am trying to generate a line graph. I am using four variables. My outcome, which would go on the y axis, is binary for whether an individual is employed or not. The x axis variable is the total number of siblings. I am trying to graph the percentage of individuals employed at every value of number of children by four groups: mother (of individual) employed, mother unemployed, father employed, father unemployed.

My outcome y axis variable is cremp2, my x axis variable is totalsib and the two variables for the groups are mtempst and ftempst.

I have tried following the advice in this post: https://www.statalist.org/forums/for...alues-by-group ; but without success.


Please see below an example of my dataset

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte cremp2 float totalsib byte(ftempst mtempst) int round
0  3 1 0 2012
0  2 1 0 2018
0  6 1 0 2012
0  7 1 0 2018
1  3 1 0 2012
1  4 0 0 2018
0  3 1 0 2012
0  3 0 0 2018
0  1 1 1 2012
0  2 1 0 2018
0  6 1 0 2012
0  6 1 0 2018
1  3 1 0 2012
0  3 1 0 2018
0  3 1 0 2012
0  3 1 1 2018
0  4 1 0 2012
0  5 1 0 2018
0  5 1 0 2012
0  4 1 0 2018
0  6 1 0 2012
0  3 1 0 2018
0  3 1 0 2012
0  2 1 0 2018
0  5 1 0 2012
0  7 1 0 2018
0  5 1 0 2012
0  4 1 0 2018
0  5 1 0 2012
1  1 0 1 2018
0  5 1 0 2012
0  5 1 0 2018
1  2 1 1 2012
1  2 1 0 2018
0  5 1 0 2012
0  5 1 0 2018
0  8 1 0 2012
1  7 1 0 2018
0  5 1 0 2012
0  4 1 0 2018
0  5 1 1 2012
0  5 1 1 2018
0  3 1 1 2012
0  3 1 1 2018
0  2 1 0 2012
0  2 1 0 2018
0 11 1 0 2012
0 12 1 0 2018
0  5 1 0 2012
1  5 1 0 2018
1  3 1 1 2012
0  3 1 0 2018
0  3 1 0 2012
1  3 1 0 2018
0  5 1 1 2012
0  5 1 0 2018
0  4 1 0 2012
1  4 1 0 2018
0  6 1 0 2012
0  7 1 0 2018
0  3 1 0 2012
1  3 1 0 2018
0  5 1 0 2012
0  5 1 0 2018
0  6 1 0 2012
0  5 1 0 2018
1  5 1 0 2012
1  5 1 0 2018
0  4 1 0 2012
1  4 1 0 2018
0  5 1 0 2012
0  5 0 0 2018
0  2 1 0 2012
0  2 0 0 2018
0  3 1 0 2012
0  4 1 0 2018
0  3 1 0 2012
0  4 1 0 2018
0  5 1 0 2012
0  6 1 0 2018
0  3 1 0 2012
0  5 1 0 2018
0  5 1 0 2012
0  5 1 0 2018
0  6 1 0 2012
0  6 1 0 2018
0  5 1 0 2012
0  5 1 0 2018
0  5 1 0 2012
0  5 1 0 2018
0  4 1 1 2012
0  4 1 1 2018
1  2 1 1 2012
1  2 1 0 2018
1  2 1 0 2012
1  1 1 0 2018
0  4 1 0 2012
0  4 0 0 2018
0  7 1 0 2012
0  7 1 0 2018
end
label values cremp2 emp
label def emp 0 "No", modify
label def emp 1 "Yes", modify
label values ftempst femp
label def femp 0 "Unemployed", modify
label def femp 1 "Employed", modify
label values mtempst memp
label def memp 0 "Unemployed", modify
label def memp 1 "Employed", modify

EDIT:

I have made some progress with the following code:

egen crempmean = mean(cremp2), by(totalsib ftempst)
egen crmean = mean(100 * cremp2), by(totalsib ftempst)
twoway connected crmean totalsib , sort by(ftempst) saving (g1)
 
egen crempmean1 = mean(cremp2), by(totalsib mtempst)
egen crmean1 = mean(100 * cremp2), by(totalsib mtempst)
twoway connected crmean1 totalsib , sort by(mtempst) saving (g2)

gr combine g1.gph g2.gph

Which produces this graph - but is there anyone to get them all onto one graph?

[Nick Cox]

Your variables ftempst mtempst jointly define four categories, 0 0, 1 0, 0 1 and 1 1. That being so, those variables cut across each other and I can't see a way to combine the graphs. without sacrificing something.

[Sherine Maui]

I see what you mean Nick, thanks. I'll keep them as separate graphs then.

[Nick Cox]

You could go for a composite. Here's the whole code I tried, starting with your data example. Presumably the graphs are more interesting for your full dataset.

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte cremp2 float totalsib byte(ftempst mtempst) int round
0  3 1 0 2012
0  2 1 0 2018
0  6 1 0 2012
0  7 1 0 2018
1  3 1 0 2012
1  4 0 0 2018
0  3 1 0 2012
0  3 0 0 2018
0  1 1 1 2012
0  2 1 0 2018
0  6 1 0 2012
0  6 1 0 2018
1  3 1 0 2012
0  3 1 0 2018
0  3 1 0 2012
0  3 1 1 2018
0  4 1 0 2012
0  5 1 0 2018
0  5 1 0 2012
0  4 1 0 2018
0  6 1 0 2012
0  3 1 0 2018
0  3 1 0 2012
0  2 1 0 2018
0  5 1 0 2012
0  7 1 0 2018
0  5 1 0 2012
0  4 1 0 2018
0  5 1 0 2012
1  1 0 1 2018
0  5 1 0 2012
0  5 1 0 2018
1  2 1 1 2012
1  2 1 0 2018
0  5 1 0 2012
0  5 1 0 2018
0  8 1 0 2012
1  7 1 0 2018
0  5 1 0 2012
0  4 1 0 2018
0  5 1 1 2012
0  5 1 1 2018
0  3 1 1 2012
0  3 1 1 2018
0  2 1 0 2012
0  2 1 0 2018
0 11 1 0 2012
0 12 1 0 2018
0  5 1 0 2012
1  5 1 0 2018
1  3 1 1 2012
0  3 1 0 2018
0  3 1 0 2012
1  3 1 0 2018
0  5 1 1 2012
0  5 1 0 2018
0  4 1 0 2012
1  4 1 0 2018
0  6 1 0 2012
0  7 1 0 2018
0  3 1 0 2012
1  3 1 0 2018
0  5 1 0 2012
0  5 1 0 2018
0  6 1 0 2012
0  5 1 0 2018
1  5 1 0 2012
1  5 1 0 2018
0  4 1 0 2012
1  4 1 0 2018
0  5 1 0 2012
0  5 0 0 2018
0  2 1 0 2012
0  2 0 0 2018
0  3 1 0 2012
0  4 1 0 2018
0  3 1 0 2012
0  4 1 0 2018
0  5 1 0 2012
0  6 1 0 2018
0  3 1 0 2012
0  5 1 0 2018
0  5 1 0 2012
0  5 1 0 2018
0  6 1 0 2012
0  6 1 0 2018
0  5 1 0 2012
0  5 1 0 2018
0  5 1 0 2012
0  5 1 0 2018
0  4 1 1 2012
0  4 1 1 2018
1  2 1 1 2012
1  2 1 0 2018
1  2 1 0 2012
1  1 1 0 2018
0  4 1 0 2012
0  4 0 0 2018
0  7 1 0 2012
0  7 1 0 2018
end
label values cremp2 emp
label def emp 0 "No", modify
label def emp 1 "Yes", modify
label values ftempst femp
label def femp 0 "Unemployed", modify
label def femp 1 "Employed", modify
label values mtempst memp
label def memp 0 "Unemployed", modify
label def memp 1 "Employed", modify

* NJC starts here

gen group = mtempst + 2 * ftempst
label def group 0 neither 1 father 2 mother 3 both
label val group group

egen mean = mean(cremp2), by(totalsib group)
separate mean, by(group) veryshortlabel

twoway connected mean? totalsib , sort scheme(s1color) ms(Oh Dh Th Sh) mc(black blue magenta red) lc(black blue magenta red) legend(col(1) pos(1) ring(0)) xla(1/12) name(G1)

twoway connected mean? totalsib, by(group, legend(off) note("")) sort scheme(s1color) ms(Oh Dh Th Sh) mc(black blue magenta red) lc(black blue magenta red) xla(1/12) name(G2)

EDIT: I think I have "father" and "mother" wrong way round here, but hope the principle is clear.

[Sherine Maui]

Thank you Nick, these graphs are what I was trying to produce and they are indeed more informative using my full sample.