Running Multiple Regressions using the Same Independent Variables

[Mark Weiss]

Hi all,

I have 30 different regressions I would like to run using the same independent variables. I understand that when I set up locals as follows, independent variables are easy to reference during the regression (let y1 be some outcome of interest):

local controls x1 x2 x3 i.x4
reg y1 $controls

However, when I define a local with a set of outcome variables, I cannot get it to work in a loop:

local outcomes y1 y2
foreach i in outcomes {
    reg $outcomes $controls
}

The above only produces one regression that has y1 as the dependent variable and y2 as an independent variable (the controls work). Does anybody know what I'm doing wrong here? Also, of secondary concern, I would like to store each estimate as the name of the dependent variable so that I can visualize the results with

estimates table y1 y2, se p keep(coefficient of interest)

[Andrew Musau]

You are looping over outcomes, so you want

local controls x1 x2 x3 i.x4
local outcomes y1 y2
foreach outcome of local outcomes{
    reg `outcome' `controls'
    est sto `outcome'
}

Also note that you are mixing up locals and globals in #1. Locals are referenced with single quotes as above, globals with dollar signs.

[Mark Weiss]

Hey Andrew,
Thanks for the response! And apologies for the global/local mixup, I originally tried this with globals instead and forgot to change my code.

I ran the loop and the .do file runs without producing any output. When I run estout `outcomes', the console tells me "last estimates (.) not found." If I run estout y1, for example, the console tells me "estimation result y1 not found." Any idea what's producing these errors?

[Andrew Musau]

Do you have the dataset loaded? It's hard to say without seeing the actual Stata output. Copy and paste this within CODE delimiters.

[Mark Weiss]

Here's what happens:

local controls i.year i.statename female black asian hispanic age65yr college hs own_home employed

. local outcomes coverage persdoc2

.
. foreach outcome of local outcomes {
  2.         reg `outcome' `controls' [aw = weight], cluster(statename)
  3.         est sto `outcome'
  4. }

.
. estout coverage
estimation result coverage not found

[Andrew Musau]

I do not see any commands loading the dataset, e.g.,

use mydata, clear

You need to make sure that the dataset is loaded and includes all the variables specified. Also, why not run this directly by copying and pasting into the command window? Any problems will become evident in this way.

[Mark Weiss]

Hey Andrew - coming back to this thread to say that your method worked once I explicitly wrote a command to load the data. Thank you so much for your help!

[roberto kimura]

You are looping over outcomes, so you want

local controls x1 x2 x3 i.x4
local outcomes y1 y2
foreach outcome of local outcomes{
reg `outcome' `controls'
est sto `outcome'
}

Also note that you are mixing up locals and globals in #1. Locals are referenced with single quotes as above, globals with dollar signs.

Hi Andrew,

is it possible to run the regressions but different regressors and regressands simultaneously?

e.g., reg1: y1 x1 x2, reg2: y1 x1 x2 x3, reg3: y2 x1 x2 etc etc

thanks in advance!

[Felix Bittmann]

Yes, try

help sureg