Adding prefix to each element in a global

[Emil Alnor]

I have a long list of variable names stored in a global. I use these to perform a series of regressions in models excluding and including control variables. I want to plot the coefficients but don't want to write out each coefficient name when using -coefplot. For simplicity I just use three variable names in the example:

global y y1 y2 y3
foreach y of varlist $y{
reg `y' x
estimates store naiv_`y'
reg `y' x z1 z2 z3
estimates store cov_`y'
}

Now my coefficients are stored with the names

naiv_y1 naiv_y2 naiv_y3 cov_y1 cov_y2 cov_y3

so writing

coefplot $y

won't do, because I need the prefixes 'naiv_' and 'cov_'. So how do I make two new globals, which are the elements in $y with the prefix 'naiv_' and 'cov_', which I then can use the plot the coefficients from the different models? Ultimately I would like to end up with two plots, 1) containing the coefficients from all the models excluding control variables, 2) contatining the coefficients from all models including control variables.

[Nick Cox]

One answer is never to use (your own) globals without good reason, and Clyde Schechter will affirm that there is never a good reason.

I don't follow all of #1 but here is some technique

foreach y in y1 y2 y3 {
    
reg `y' x
estimates store naiv_`y'
local wanted1 `wanted1' naiv_`y'
reg `y' x z1 z2 z3
estimates store cov_`y'
local wanted2 `wanted2' cov_`y'

}

You may seek the concatenation

local wanted `wanted1' `wanted2'

[Emil Alnor]

Nick Cox I started using globals since the confusion my use of locals caused to me here https://www.statalist.org/forums/for...r-the-variable. That is, I don't see how globals can cause similar errors, because they can work even when running code one line at a time. Not that I am strongly convinced that globals don't have their weaknesses, I just haven't experienced them yet in my use of Stata.

Anyway, thanks for the code!