# Re: st: A question about bootstrap

 From "Alfonso Miranda Caso Luengo" To Subject Re: st: A question about bootstrap Date Tue, 10 Aug 2004 16:03:06 +0100

```Dear Tak-wai Chau,

The first question I would ask is whether bootstrapping standard errors from your ml rutine will be practical. Notice that bootstrapping needs running your ml code many times. With large data sets and slow ml routines this could be an important issue. Since you have cross section data it seems that you can use the lf method and that helps speed a lot.

Besides your ml code you will need to write a small program to perform the bootstrapping. This is a minor problem as it is fairly easy if you know the basics of Stata programming. Have a look in the manuals.

If I remember well you can adjust the covariance matrix for the first stage estimation and it should be simple. Have a look on chap. 16 and 17 in Greene's Econometric Analysis.

Best,

Alfonso.

=====================================
Alfonso Miranda
PhD Student

Economics Department
University of Warwick
Coventry CV4 7AL
E-mail:ecrgw@warwick.ac.uk
WebPage: http://www.warwick.ac.uk/go/amiranda
=====================================
>>> tchau@troi.cc.rochester.edu 08/10/04 15:18 PM >>>
Dear all,

I have some questions in using bootstrap in my task.

I would like to estimate labor supply equations of husbands and wives.
(One for husbands, one for wives) I am going to use a two-step
procedure. First, I estimate the husband's and wife's wage equations on
instruments by OLS. Second, I estimate jointly the husband's and wife's
labor supply equations with Maximum Likelihood (using the predicted
wages in the first step). For this step, I am going to use the ml commands.

I would like to use bootstrap to obtain the standard errors or
confidence interval of parameters of the labor supply equation (reduced
form) and some non-linear functions of these parameters. I want to do
this because I don't know how to adjust standard error for the first
stage estimation, and also approximation for standard errors may be poor
for non-linear functions of parameters.

The data is from a cross-sectional data so IID can be assumed.

I have the following questions in actually carrying out the above
estimation.

1. Is my understanding of the use of bootstrap correct? Any special
thing I should take care in my case?
2. I know there is a bootstrap command in Stata. Can it be used to do
bootstrap in a two-stage process?
3. How can I obtain the standard error/confidence interval estimates
from bootstrap of a non-linear function of parameter in Stata?

Thank you very much in advance!

Regards,
Tak-wai Chau

*
*   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/
```