How to make coefplot treat coefficients from different models that have the same regex as the same

[Arthur Carvalho Brito]

Hello all, could someone help with the following, please?

Suppose I run two regressions as follows:

eststo m1: reghdfe y  x1_1 x2_1
eststo m2: reghdfe y  x1_2 x2_2

I want then to plot these two estimates using coefplot. However, variables x1_1 and x1_2 should be considered the same regressor, likewise for x2_1 and x2_2. That is, my ideal graph would have labels x1 and x2 on the y axis, with 2 models associated to each.

How can I do that? The estimation procedure cannot be altered, as it is ran in a for loop beforehand and all variables belong to the same dataset.

I tried using something like

rename(x1_* ="x1" x2_* ="x2", regex) ///

in the coefplot call, with no results

[Luca Calianno]

This is just a simple example but I hope it gives you a possible intuition on how to do it.

* Here I just create a dataset similar to yours
clear 
set obs 1000

gen x1_1 = runiform()
gen x2_1 = runiform()
gen x1_2 = x1_1
gen x2_2 = x2_1

gen y = x1_1 + x2_1 + rnormal()


* Here I run the regression and store the results in a new matrix called beta1 
eststo m1: reghdfe y  x1_1 x2_1
matrix beta1= e(b)
matrix colnames beta1 = "x1" "x2"
estadd matrix beta = beta1: m1

* Same for model 2
eststo m2: reghdfe y  x1_2 x2_2
matrix beta2 = e(b)
matrix colnames beta2 = "x1" "x2"
estadd matrix beta = beta2: m2

coefplot m1 m2, b(beta)