Colour Coding Markers In a Weighted Scatter Plot

[vincent fernandes]

I am trying to create a scatter plot of Disease Mortality (CD_Mortality) and Income (GDP) weighted by Health Expenditure (Health_Exp), I have being able to produce this using the following code:

twoway (scatter CD_Mortality GDP [aweight = Health_Exp], mcolor(navy8) msymbol(smcircle_hollow))

I want to colour code each country with respect to their WHO region indicator (WHO_IND), however if I separate CD_Mortality into the regions and plot using the twoway scatter function it will weight each category individually (See attachment). Is there a way for the marker colour to represent the WHO region and for the marker to be weighted across all observations not just its WHO region. Thank you for your time. (Using STATA 16)

[Andrew Musau]

I saw this post a couple of days ago, but I didn't bother to attempt a solution since you do not provide a data example. Please review the FAQs on how to pose questions to increase your chances of getting a helpful reply. Here, the issue is that your weights are not consistent across groups whereas Stata weights markers within groups. There is an easy fix to this... the idea can be found in this thread.

Reproducible example:

sysuse auto, clear
*SCATTER MPG WEIGHT WEIGHTED BY REP78
twoway (scatter mpg weight [aweight = rep78], mcolor(navy8) ///
msymbol(smcircle_hollow)), scheme(s1color)
gr save gr1.gph, replace

*DIFFERENTIATE GROUPS BY FOREIGN
twoway (scatter mpg weight [aweight = rep78] if foreign, mcolor(navy8) ///
msymbol(smcircle_hollow))(scatter mpg weight [aweight = rep78] if ///
!foreign, mcolor(blue) msymbol(smcircle_hollow)), scheme(s1color)
gr save gr2.gph, replace
gr combine gr1.gph gr2.gph, scheme(s1color)

Here, you can see that with inconsistent weights, the markers across groups have different sizes (see red arrows in the attached figure). You can make sure that weights are consistent across groups by duplicating observations and switching categories for the duplicated set. Thereafter, replace observations of the variables that you use with missing values.

expand 2, gen(set)
recode foreign (0=1) (1=0) if set
foreach var in mpg weight{
replace `var'=. if set
}
twoway (scatter mpg weight [aweight = rep78] if foreign, mcolor(navy8) ///
msymbol(smcircle_hollow))(scatter mpg weight [aweight = rep78] if ///
!foreign, mcolor(blue) msymbol(smcircle_hollow)), scheme(s1color)
gr save three.gph, replace
gr combine one.gph two.gph three.gph, scheme(s1color)

[Jessica Pan]

Hello, any suggestions on how to do this when there are more than two groups? For example, based on the previous examples, the following code works well for 2 groups:

Code:
clear all
input x y weight group
1 1 1 1
2 1 2 1
1 2 4 2
2 2 8 2
end

expand 2, gen(set)
recode group (2=1) (1=2) if set
foreach var in y x{
replace `var'=. if set
}
twoway (scatter y x if group==1 [w=weight]) (scatter y x if group==2 [w=weight]), legend(off)

But what if there are four groups? For example using the data below?

Code:
clear all
input x y weight group
1 1 1 1
2 1 2 1
1 2 4 2
2 2 8 2
1 3 10 3
2 3 12 3
1 4 14 4
2 4 18 4
end

[Andrew Musau]

I have subsequently worked out a general solution to this problem and based on my example in #2 is

expand 2, gen(set)
replace foreign=-1 if set
fillin foreign rep78
drop if set==1

provided that

tab foreign

does not include a category with the value -1.