Re: st: Conditional matching of observations

[Christopher Hajzler <chajzler@gmail.com>]

here

On Thu, Mar 4, 2010 at 6:15 PM,  <sjsamuels@gmail.com> wrote:
> The technical answer to your question is simple: give every person in
> A (receive support) a pseudo-age and and gender which you will use to
> match to people in B (do not receive support). Then use one of Stata's
> many matching programs to match on race, pseudo-age, and
> pseudo-gender.
>
> Female in A: pseudo-age = age +3 pseudo-gender = male
> Male in A:  pseudo-age = age-3 pseudo-gender = female
>
> Female in B:  pseudo-age = age pseudo-gender =female
> Male in B:  pseudo-age = age pseudo-gender = male
>
>
> However I question your basic design. Even if the average age of males
> paying child support to females is three years older than the females
> they support, it is not exactly three years older in every case. Your
> matched population of potential payers  will match the actual
> population only in mean difference in age but not in any other
> characteristics, such as the quantiles of the difference. Also the
> population B who do not receive child support will include a
> supopulation, B', approximately equal in size to A, who already pay
> child support to someone in A.   You are also ignoring geographical
> proximity, which is important for many phenomena. Will ignoring these
> issues cause problems? Without knowing the purpose and details of your
> study it is difficult to say much more.
>
> Steve
> On Wed, Mar 3, 2010 at 6:08 PM, Nathan Hutto <nathandavidhutto@gmail.com> wrote:
>>
>> I have a large population data set. I want to match people who have
>> received child support payments (0/1) to a person who could have
>> theoretically paid them child support. I want to match on sex, race,
>> and age only. For each person who receives child support, I want to
>> match to someone of the opposite sex and same race. If the person
>> receiving is a female, her match should be 3 years older. If the
>> person receiving the match is a male, his match should be 3 years
>> younger. I have no idea how to go about do this after searching high
>> and low for an appropriate method. Any help would be much appreciated.
>>
> *
> *   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/