Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: warning: overlapping samples


From   Steven Samuels <[email protected]>
To   [email protected]
Subject   Re: st: warning: overlapping samples
Date   Sat, 30 Jun 2007 22:08:58 -0400

Dear [email protected]:

Ben advised you to use the bootstrap to compute standard errors. Here is how to do it.

First, put your data into "long" form, with one line for each person. You will need a pair "id" to identify married couples. Data from two couples might look like:

pair_id gender wages x1 x2
1 male 10000 13 6
1 female 5000 9 6
2 male 8000 12 5
2 female 40000 7 19

If you have single people, give each a unique pair_id.

Assume you plan to regress wages on x1 and x2. To find the oaxaca decomposition of the male-female differences, use this code:

/*********************************************************************/
set seed 498733080 //substitute your choice of random number seed
bootstrap _b, cluster(pair_id): oaxaca2 wages x1 x2, by(gender) esave nose
/**********************************************************************/

Substitute your own "x" variables for "x1 x2". Ben's command - oaxaca2- is a version of -oaxaca- that he designed for use with the bootstrap.

I don't know this field. But according to one paper, it is common to use log(wages) as the dependent variable. See: http:// www.ssc.govt.nz/display/document.asp?DocID=4692

One more thing: Statalist is a friendly forum. It is common practice to sign at least a first name to email. I hope that you will do this when you reply.

Good luck!

-Steve

On Jun 28, 2007, at 11:30 AM, Ben Jann wrote:


You can use -oaxaca- in this case, but the standard errors will be
biased because the sample of females and males are not independent.
Use the bootstrap to compute the standard errors.
ben

On 6/28/07, [email protected] <[email protected]> wrote:
Thanks Ben, I don't know why? because I want calculate the
decoposition between husband and wife, so sure I have relation because
my dataset is:
wife husband wwage hwage
1 1 100 200
I do :
reg wwage xi
est stor w
reghwage xi
est sto h
oaxaca h w


Can I estimate with oaxaca this data set? is it right the procedure?
thanks a lot for your help.



Quoting Ben Jann <[email protected]>:

> It means that some of your females are in the male sample or vice
> versa. Check the -if- qualifiers of your regression commands.
> ben
>
> On 6/28/07, [email protected] <[email protected]> wrote:
>> Dear statalist, I'm doing a oaxaca decomposition and I have this
>> message before the estmation with oaxaca. What is it means?
>> thanks
>> warning: overlapping samples
>> (high estimates: female4; low estimates: male4)
>>
>> Mean prediction 1
>> = 2.095566
>> Mean prediction 2
>> = 1.939635
>> --------------------------------------------------------------------- ---------
>> | Coef. Std. Err. z P>|z| [95%
>> Conf. Interval]
>> ------------- +----------------------------------------------------------------
>> difference | .155931 .0301257 5.18 0.000 .0968858
>> .2149762
>> --------------------------------------------------------------------- ---------
>>
>> Linear decomposition
>> --------------------------------------------------------------------- ---------
>> Total | Coef. Std. Err. z P>|z| [95%
>> Conf. Interval]
>> ------------- +----------------------------------------------------------------
>> three-fold |
>> endowments | -.6220027 .2610409 -2.38 0.017 -1.133633
>> -.110372
>> coefficients | -.1470742 .0839665 -1.75 0.080 -.3116456
>> .0174971
>> interaction | .9250079 .2727451 3.39 0.001 .3904373
>> 1.459579
>> --------------------------------------------------------------------- ---------
>>
>>
>>
>> *
>> * For searches and help try:
>> * http://www.stata.com/support/faqs/res/findit.html
>> * http://www.stata.com/support/statalist/faq
>> * http://www.ats.ucla.edu/stat/stata/
>>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/




*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index