esttab drop option

[Daniel Feenberg]

I have many similar regressions (in a loop) with many nuisance variables. As the sample changes, sometimes dummies are dropped from the regression by the xtreg command itself.. If I have listed the variable in the -esttab- -drop- optiion, that is a fatal error. I would prefer not to have hundreds of meaningless insignificant coefficients cluttering the output. Is there a way around this, or is there a replacement for -esttab- that avoids this problem? I am using version 2.0.5 from SSC.

Daniel Feenberg
NBER
http://www.nber.org/stata/efficient

[Roberto Ferrer]

Assume you have a setup similar to this (in a loop):

eststo clear
        
sysuse auto, clear

// fixed list of vars to drop
local todrop mpg

// regression 1       
quietly regress price weight mpg rep78 foreign
esttab, ar2 drop(`todrop')

// regression 2        
quietly regress price weight foreign
esttab, ar2 drop(`todrop')

which terminates in this error:

coefficient mpg not found
r(111);

Then consider something like:

eststo clear
        
sysuse auto, clear

// fixed list of vars to drop
local todrop mpg

// regression 1       
quietly regress price weight mpg rep78 foreign

local all = e(cmdline)
local cmddepvar = "`e(cmd)' `e(depvar)'"
local indepvar: list all - cmddepvar
local tokeep: list indepvar - todrop

esttab, ar2 keep(`tokeep')

// regression 2        
quietly regress price weight foreign

local all = e(cmdline)
local cmddepvar = "`e(cmd)' `e(depvar)'"
local indepvar: list all - cmddepvar
local tokeep: list indepvar - todrop

esttab, ar2 keep(`tokeep')

See -help macrolists-. Also http://www.stata.com/statalist/archi.../msg00369.html for other options on recuperating the list of regressors.

[Daniel Feenberg]

Thanks for the suggestion. I guess I was hoping I had missed something simple.

I have one more problem with -esttab-, though. If I have

eststo m1 : regress y x
esttab eqlabels("m1")

shouldn't the output column be headed with "m1" rather than "(1)"? I have found several examples for the eqlabel option on the Stata website and elsewhere, but haven't found any documentation for it. Is there documentation that would tell me what I was doing wrong?

Daniel Feenberg
NBER