How can we get the names of the independent variables in the regression like e(depvar) for later use?

[Kazuto Sumita]

Dear Stata users:

Hello!
I'd like to ask you the way to get the names of the independent variables in the "regress" command for the post-use.
Regarding the dependent variable, I know there is e(depvar).
However, are there any useful command for independent variables?

Thanks in advance:
Kazuto

[Richard Williams]

There is probably a simpler way, but maybe something like

webuse nhanes2f, clear
reg weight height i.female
local depvars regress `=e(depvar)'
local indvars: list cmdline - depvars
macro list _indvars

[Clyde Schechter]

I think Richard forgot to include a line in his code:

webuse nhanes2f, clear
reg weight height i.female
local depvars regress `e(depvar)'
local cmdline `e(cmdline)'
local indvars: list cmdline - depvars
macro list _indvars

[Richard Williams]

Clyde is right. The local macro cmdline had been previously defined by me so I didn't get an error when I ran my earlier code.

Incidentally, all sorts of things will cause even the corrected code to fail, e.g. specifying options as well as variables. But presumably you could come up with an error-prone approach.

[Kazuto Sumita]

Thanks Richard and Clyde!
Your comments really help me a lot!

Best Regards

Kazuto

[Richard Williams]

Turns out Maarten Buis wrote a program for this 10 years ago. Type -findit indeplist-.

[Kazuto Sumita]

Thanks a lot!
This command is useful in the case that we have conditions and group dummies.

xi : reg y x  i.area if year==2015
indeplist
local x = r(X)

Hear area is the variable indicating the group.

Kazuto

[Richard Williams]

Unless you have a horribly antiquated version of Stata, I wouldn't use xi:. Factor variables are usually better. See -help fvvarlist-.

[Kazuto Sumita]

Thanks for your comment!
However, "area" is a string variable,
such as "area1, area2,..."
And without "xi:", I obtained the following error message:

reg y x  i.area if year==2015

area:  string variables may not be used as factor variables
r(109);

How can I cope with dummies created from string variables?

[Jeph Herrin]

Using -c(cmdline)- can get messy if the command is not -regress-; for instance

mixed y x weight || id:, vce(robust)


means having to parse out things that aren't variable names.

The best solution is to store the list in advance, since you are typing them anyway:

local ivars x weight
mixed y `ivars' || id:


but if some reason that's not practical, I typically use the column names from e(b); there will still be some things that aren't variable names, but it's easy to loop over them and -confirm variable- or such.

hth,
Jeph

[Kazuto Sumita]

Dear Jeph

Thank you!

Best,
Kazuto