How do I change the color of the text in my legend, in twoway?

[Rory Todd]

My colleague created a graph in Stata using the code below. I'd like to make the writing "By eliminating lead" in the legend appear blue, like the writing of 'hlo_expost1'. How do I achieve this? Thanks!
#delimit ; tw (pccapsym rank hlo rank hlo_expost1, msym(none) lcolor(blue)) (sc rank hlo, mlabel(hlos1) msymbol(none) mlabpos(9)) (sc rank hlo_expost1, mlabel(effect1s1) msymbol(none) mlabpos(3) mlabc(blue)) (sc rank left, mlabel(countryname) msymbol(none) mlabpos(3) mlabcolor(black)), ysc(off) xsc(range(650)) legend(order(2 "Status quo" 3 "By eliminating lead")) ylab(,nogrid) xlab(0(100)600,nogrid) xline(500) xsize(4) ysize(6) xtitle("World Bank, Harmonized Learning Outcomes", margin(20 0 0 0) height(5)) ; #delimit cr

[Nick Cox]

Cross-posted at https://stackoverflow.com/questions/...n-stata-twoway

[Andrew Musau]

help legend options

webuse grunfeld, clear
set scheme s1mono
tw line invest year if company==1|| line invest year if company==2, ///
sort leg(order(1 "some text" 2 "some other text") color(red))

[Hemanshu Kumar]

Andrew Musau I think the challenge in OP's problem is to have different colours for the text of different legend keys.

[Nick Cox]

#4 That was my reading too.

[John Mullahy]

I can't figure out a legend option in a twoway command that does this, but graph editor seems to be up to the task.

[Andrew Musau]

Andrew Musau I think the challenge in OP's problem is to have different colours for the text of different legend keys.

You are correct, I did not read the question properly. As the legend is considered one entity, I think the only way to do it is through the graph editor as suggested by Nick and John.

[Nick Cox]

You could use text() options to put separate text in different places with different colours.

This example uses stc1 and stc2 for text items in Stata 18 matching default line colours, but in any other version since Stata 8 you just need to spell out the colours you want.

webuse grunfeld, clear
su invest if year == 1954 & company == 1, meanonly
local last1 = r(max)
su invest if year == 1954 & company == 2, meanonly
local last2 = r(max)
tw line invest year if company==1 || line invest year if company==2, ///
sort text(`last1' 1954.2 "1", color(stc1)) text(`last2' 1954.2 "2", color(stc2)) legend(off)

Such direct labelling (labeling) is in many ways often preferable to a legend any way.

[Sam Lindquist]

The problem with editing the legend within the graph editor is that it also changes the colors of the symbols in the graph itself. Im creating a graph with two different groups, using symbol shapes to indicate membership of one group and colors to indicate membership of another. Thus, in the graph, every point in the scatter plot has some color, but I want the shapes in the legend to all show up as black. It seems like it cant be done...

Code:

twoway (scatteri 10 10, mcolor(%0)) /// * graphing invisible point that will be used as the title of the 
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "No Goal" & practicegroup == "Easy", lc(red)) /// * first graphing se's
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "Attainable" & practicegroup == "Easy", lc(red)) ///
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "Baseline" & practicegroup == "Easy", lc(red)) ///
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "Stretch" & practicegroup == "Easy", lc(red)) ///
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "No Goal" & practicegroup == "Medium", lc(blue)) ///
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "Attainable" & practicegroup == "Medium", lc(blue)) ///
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "Baseline" & practicegroup == "Medium", lc(blue)) ///
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "Stretch" & practicegroup == "Medium", lc(blue)) ///
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "No Goal" & practicegroup == "Hard", lc(green)) ///
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "Attainable" & practicegroup == "Hard", lc(green)) ///
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "Baseline" & practicegroup == "Hard", lc(green)) ///
||       (rcap num_tables_coef_value_max num_tables_coef_value_min goal_coef_value if goallevel == "Stretch" & practicegroup == "Hard", lc(green)) ///
||     (scatteri 10 10, mcolor(%0)) ///
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "No Goal" & practicegroup == "Easy", mcolor(red) msymbol(S)) /// * now graphing point estimates
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "Attainable" & practicegroup == "Easy", mcolor(red) msymbol(O)) /// 
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "Baseline" & practicegroup == "Easy", mcolor(red) msymbol(T)) ///
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "Stretch" & practicegroup == "Easy", mcolor(red) msymbol(X)) ///
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "No Goal" & practicegroup == "Medium", mcolor(blue) msymbol(S))  ///
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "Attainable" & practicegroup == "Medium", mcolor(blue) msymbol(O)) ///
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "Baseline" & practicegroup == "Medium", mcolor(blue) msymbol(T)) ///
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "Stretch" & practicegroup == "Medium", mcolor(blue) msymbol(X)) ///
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "No Goal" & practicegroup == "Hard", mcolor(green) msymbol(S)) ///
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "Attainable" & practicegroup == "Hard", mcolor(green) msymbol(O)) ///
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "Baseline" & practicegroup == "Hard", mcolor(green) msymbol(T)) ///
||       (scatter num_tables_coef_value goal_coef_value if goallevel == "Stretch" & practicegroup == "Hard", mcolor(green) msymbol(X)) ///
||     (function y = x, range(0 15) lc(grey%30)), ///
legend(order(1 "Practice Difficulty" 2 "Easy" 6 "Medium" 10 "Hard" 14 "Goal Level" 15 "No Goal" 16 "Attainable" 17 "Baseline" 18 "Stretch") ///
        holes(5) ///
        symxsize(3) ///
        rows(10) ///
        cols(1) ///
        colfirst ///
        notextfirst ///
        position(3) ///
) ///
xlabel(, labsize(medlarge)) ///
ylabel(, labsize(medlarge)) ///
xtitle("Goal", size(large)) ///
ytitle("Number of Tables Completed", size(large))

[Andrew Musau]

The problem with editing the legend within the graph editor is that it also changes the colors of the symbols in the graph itself. Im creating a graph with two different groups, using symbol shapes to indicate membership of one group and colors to indicate membership of another. Thus, in the graph, every point in the scatter plot has some color, but I want the shapes in the legend to all show up as black. It seems like it cant be done...

This thread deals with the legend text, not symbols. The colors of the symbols in the legend correspond to the colors of symbols in the graph itself. Previously, there have been discussions on how to include legend symbols that do not appear in the graph (see, e.g., #7 https://www.statalist.org/forums/for...-in-the-legend). You may consider a different colored symbol as a separate symbol in its own right.

[Sam Lindquist]

Hi Andrew, I checked out that other thread and solved the issue. Thanks!