RE: st: Creating dummy variables

[Michael Betz <betz.40@buckeyemail.osu.edu>]

Thanks Matt,

This is getting close but there is still a hang-up. The program you wrote differences all "fips1" dummies with all "fips2" dummies. I need to get difference dummies only for the pairs (i.e 1001-1073, 1001-1021, and 1001-1101, but not 1001-12031). Because I have 3,000 levels for each "fips" variable, this program would create 3,000 x 3,000 variables, which is where Stata runs into a problem. 

Mike
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Matthew White
Sent: Wednesday, November 16, 2011 9:14 AM
To: statalist@hsphsun2.harvard.edu
Subject: Re: st: Creating dummy variables

Hi Mike,

There's probably a more efficient way to do this, but here's one way:

***BEGIN***
clear
input fips1 fips2
1001 1073
1001 1021
1001 1101
1003 12031
1003 1099
end

forvalues i = 1/2 {
	levelsof fips`i'
	foreach county in `r(levels)' {
		generate fips`i'_`county' = fips`i' == `county'
		label variable fips`i'_`county' "fips`i'==`county'"
		
		local dummies`i' `dummies`i'' fips`i'_`county'
	}
}

foreach dummy1 of local dummies1 {
	local num1 = substr("`dummy1'", strpos("`dummy1'", "_") + 1, .)
	
	foreach dummy2 of local dummies2 {
		local num2 = substr("`dummy2'", strpos("`dummy2'", "_") + 1, .)
		
		generate diff_`num1'_`num2' = `dummy1' - `dummy2'
		label variable diff_`num1'_`num2' "`dummy1' - `dummy2'"
	}
}
***END***

Best,
Matt

On Tue, Nov 15, 2011 at 9:44 PM, Michael Betz
<betz.40@buckeyemail.osu.edu> wrote:
> Hi all,
>
> I have two categorical variables "fips1" and "fips2" that record the US county of the observation. For each "fips1" there are many "fips2" counties as below
>
> fips1   fips2
> 1001    1073
> 1001    1021
> 1001    1101
> 1003    12031
> 1003    1099
>
> I need to create dummy variables for each county in "fips1" and "fips2" and then create variables representing the difference between the two dummy variables as below:
>
> fips1   fips2   dum1_1  dum1_2  dum2_1  dum2_2  dum2_3  dum2_4  1_1-2_1 1_1-2_2 1_1-2_3
> 1001    1003    1               0               1               0               0               0               0               1               1
> 1001    1021    1               0               0               1               0               0               1               0               1
> 1001    1101    1               0               0               0               1               0               1               1               0
> 1003    1021    0               1               0               1               0               0               0               0               0
> 1003    1001    0               1               0               0               0               1               0               0               0
>
> One added constraint is that each of "fips1" and "fips2" creates 3,000 dummies, so Stata cannot hold variables representing the difference between all pairs of dummy variables. I need to only calculate the difference in dummies for the pairs that in the data (i.e. according to the example above I would not need the difference between the dummies for "fips1"=1001 and "fips2"=1001 because that pair doesn't exist in my data)
>
> I've been thinking all day trying to come up with a solution, but to no avail. I appreciate and help or suggestions.
>
> Thanks,
> Mike
>
>
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>



-- 
Matthew White
Data Coordinator
Innovations for Poverty Action
101 Whitney Avenue, New Haven, CT 06510 USA
+1 434-305-9861
www.poverty-action.org

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