Error with user written program grc1leg. It produces 3 legends instead of 1 legend for combined 2 graphs!

[Andrea Maier]

Hi at all.

I try to use grc1leg with margins plot.


I have the strange error, that I get 3 in stead of 1 legend.

I tried to create an example, that closely follows my data, but I cannot manage with the auto dataset. But the idea is:

gen displ = 0 if displacement < = 100

replace displ = 1 if displacement > = 100

replace displ = 2 if displacement > = 200

label define displ 0 "Moederat" 1 "Medium" 2"Really high"

label values displ displ

gen expensive = 0 if price > 10000

replace expensive = 1 if price >= 10000


quietly: logit foreign c.headroom##c.weight##(1.expensive ib0.displ), ///

       noomitted noempty

  graph drop _all

quietly: margins, at(headroom=(2(0.5)4) weight=(2500) displ=(0 1 2)) over(expensive) 

marginsplot, by(expensive) scheme(s1color) name(m1) nodraw title("weight = 2.5") subtitle(, fcolor(white)) ///

 legend(cols(1) size(vsmall))  recast(line)

quietly: margins,  at(headroom =(2(0.5)4) weight=(3500)  displ=(0 1 2)) over(expensive) 

marginsplot, by(expensive) scheme(s1color) name(m2) nodraw title("weight = 3.5")subtitle(, fcolor(white)) recast(line)



grc1leg m1 m2, name(m3, replace) span legendfrom(m1)    ycommon      

graph display gme

The resulting error is in the attached picture.

Any suggestions how to overcome the problem are highly appreciated. Thanks!

[Roman Mostazir]

Its the by dimension causing the problem. I would just plot them separately in a loop and combine if `grc1leg' is needed. Try this:

gr drop _all
loc k=1
foreach j of numlist 2500 3500 {

forval i=0/1 {
    loc t : lab (expensive) `i'
    margins if expensive==`i',  at(headroom =(2(0.5)4) weight=(`j')  displ=(0 1 2))
    marginsplot,scheme(s1color) legend(size(small) col(1)) ///
     nodraw title("`t': weight = `j'") subtitle(, fcolor(white)) ///
     recast(line)
gr copy m`k', replace
loc ++k
}
}
grc1leg m1 m2 m3 m4, scheme(s1color)

[Andrea Maier]

Thanks for the answer. That works reasonably well for me, thanks! Great!