Coefplot: sort coefficients from different models

[John Kirk]

Hi all,

I've seen this question get asked a few times (e.g. here), but I have been unable to find an answer anywhere. When using coefplot to graph coefficients from multiple models, how do I sort these by the size of the point estimate?

The sort option only seems to allow sorting based on the same model/equation. Any help here would be highly appreciated!

here's a quick example:

sysuse auto, clear
*run a few models
forvalues i=3/5 {
    eststo reg_`i': reg price mpg if rep78==`i'
    }
 
*plot regression coefficient (sort option added, but as described, that doesn't do anything)
 coefplot reg_*, keep(mpg) sort()

Thanks,
John

[John Kirk]

In case this helps as a starting point I've fought around with matrices and have arrived at half a solution that lets me graph the point estimates, but they're all on the same point on the x axis and therefore I cannot distinguish CIs (see code below). Again, any help would be highly appreciated!

cap ssc install matsort
sysuse auto, clear

forvalues i=3/5 {
    eststo reg_`i': reg price mpg if rep78==`i'
    matrix test = r(table)'
    matrix test`i' = test[1,1...]

    }
matrix coefs = (test3 \ test4 \ test5)
matsort coefs 1 "down"
    coefplot (matrix(coefs[,1]))    , ci((coefs[,5] coefs[,6]))

[Andrew Musau]

coefplot is from SSC, as you are asked to explain (FAQ Advice #12). You need to be able to distinguish the coefficients and how you do this is by their names. Here is an approach from your example.

sysuse auto, clear
*run a few models
forvalues i=3/5 {
    rename * *_`i'
    eststo reg_`i': reg price mpg if rep78==`i'
    rename *_`i' *
}
set scheme s1color
coefplot reg_*, keep(mpg*) sort ylab("") ytitle("mpg", orient(horiz))

Res.:

[John Kirk]

Thanks Andrew Musau -- highly appreciated!