Scatter plot with weights and different colours

[Oliver McManus]

Hello,

I'm trying to make something like a forest plot, with markers weighted by number of observations. I also want my categories in different colours. I can't seem to combine the two: either I can specify different plots with twoway and give them different colours, but each plot is apparently weighted individually so the marker sizes are the same; or I can specify them as one plot, which allows me to use weights, but not different colours.

Here are the two things I'd like to combine:





My dataset has one observation per BMI category, with an estimate and a confidence interval for mean change per year.

This is my code at the moment, with the weighted/uncoloured version commented out:

    #delimit ;
        
    twoway    /// CIs
            (rspike ann_lci_`rfvar' ann_uci_`rfvar' bmicat if bmicat == 1
                , horizontal color($c1 )) 
            (rspike ann_lci_`rfvar' ann_uci_`rfvar' bmicat if bmicat == 2
                , horizontal color($c2 )) 
            (rspike ann_lci_`rfvar' ann_uci_`rfvar' bmicat if bmicat == 3
                , horizontal color($c3 )) 
            /// markers (colour)
            (scatter bmicat ann_est_`rfvar' if bmicat == 1 [aweight=respbmi] 
                , color($c1 ) msymbol(square) msize(medlarge)) 
            (scatter bmicat ann_est_`rfvar' if bmicat == 2 [aweight=respbmi] 
                , color($c2 ) msymbol(square) msize(medlarge)) 
            (scatter bmicat ann_est_`rfvar' if bmicat == 3 [aweight=respbmi] 
                , color($c3 ) msymbol(square) msize(medlarge))
            /* /// markers (weights)
            (scatter bmicat ann_est_`rfvar' [fweight = resp]
                , msymbol(square) msize(medlarge) mcolor(gs2))
            */
            /// options            
            , xtitle("Annual change") xline(0) xmtick(none) 
            ytitle("") yscale(range(-1 5) lcolor(none) reverse) ymtick(none)
            ylabel(1 "< 25" 2 "25-29" 3 "30+"
                , angle(horizontal) notick nogrid)
            title("Annual change in `rfname' by BMI category") 
            legend(off) scheme(s1mono) plotregion(lcolor(none)) 
            name(`rfvar'_annual_c2, replace)
            ;

    #delimit cr

I explored adding my three data points as different variables into the same scatter plot, but as you can only specify the x variable once, I could only orient my plot vertically. I looked for a way to transpose the x and y axis, but there doesn't seem to be one.

Any advice?

[Inge Olsen]

I would look into the -coefplot- package, it is very customizable.

Check this page: http://repec.sowi.unibe.ch/files/wp1...3-coefplot.pdf

Kr, Inge

[Andrew Musau]

FAQ 12 with emphasis on 12.2: You are unlikely to get a direct solution if you do not show us a sample of your data that created the graphs.