RE: st: Get list of non-omitted variables after a regression

[Joe Canner <jcanner1@jhmi.edu>]

Sorry, my solution only allows for variable names with alphabetic characters.  Since we don't have to make any assumptions about the validity of the characters following the "o." in a list of variable names, we can do the following:

. local n = regexr("`n'","o\..+","")   

(For those trying to learn about regular expressions: the "\." tells the function to look for a literal period and the ".+" tells the function to look for one or more characters after the "o.")

Note: While testing this solution on categorical variables omitted because of perfect prediction in a logistic regression, I found that the problem becomes even more complicated.  For example:

. sysuse auto, clear
. // Generate a categorical variable where one strata perfectly predicts the outcome
. gen x=recode(_n,25,50,75)
. replace x=foreign if x==50
. logistic foreign price x   // Silly example
. local n : colnames e(b)
. di "`n'"

// Output is: price 0b.x 25o.x 75o.x _cons

. local n = regexr("`n'","[0-9]+o\..+","")     // Also get rid of category values before the "o."
. di "`n'"

I'm sure there are other situations that I am missing here, so caveat emptor...

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Joe Canner
Sent: Monday, March 10, 2014 10:09 AM
To: statalist@hsphsun2.harvard.edu
Subject: RE: st: Get list of non-omitted variables after a regression

I interpreted Ludovico's question to refer to variables omitted because of collinearity or perfect prediction, but I didn't have a chance at the time to test a solution.  Building on Paul's solution, one would have to look for variables in the regression matrix that start with "o.".  For example,

. sysuse auto, clear
. gen length1=length
. regress price displacement foreign weight length length1   // Purposely putting two identical variables in so that one will be omitted
. local n : colnames e(b)
. di "`n'"
. local n = regexr("`n'","o\.[a-zA-Z]+","")   // remove any variables starting with "o."
. di "`n'"

Joe Canner
Johns Hopkins University School of Medicine
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Seed, Paul
Sent: Monday, March 10, 2014 9:47 AM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Get list of non-omitted variables after a regression

The simplest solution is to get the column names off the regression matrix.

*********** Start Stata code ***************
	sysuse auto, clear
	stepwise , pe(0.05) : regress  price mpg rep78 headroom trunk weight length turn displacement gear_ratio foreign

	local n : colnames e(b)
	di "`n'"

* If you want to lose the constant...

	local c _cons
	local n : list n-c
	di "`n'"
* Note, I am using macro extended functions to avoid silent, automatic
* truncation of macros at 245 characters whenever "=" occurs.

*********** End Stata code ******************

Paul T Seed, Women's Health, KCL


> Date: Sat, 8 Mar 2014 16:30:49 -0600
> From: Scott Merryman <scott.merryman@gmail.com>
> Subject: Re: st: Get list of non-omitted variables after a regression
> 
> See - _rmcoll , forcedrop -
> 
> Scott
> 
> 
> On Sat, Mar 8, 2014 at 3:38 PM, Ludovico Zaraga
> <ludovico.zaraga@u.northwestern.edu> wrote:
> > Hello,
> >
> > Is there a simple way of getting a list (in the form of a macro) of
> > non-omitted variables after running a regression in Stata? I am using
> > a very large number of variables, so I can't do this by hand.
> >
> > Thanks,
> >
> > Ludovico





*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/