Dear all,
I am wondering how bootstrap works with a two-step regression such as
heckman.
Suppose that I want to run the following model using bstrap.
heckman wage educ age, select(married children educ age)
Wage is missing in some cases and we suspect that there is a selection
problem.
Suppose that we have 5 observations in total
wage married
. 1
1 1
2 0
3 0
4 0
My understanding on how it works is that Bstrap would randomly select 5
observations with replacement and we may end up having a sample that looks
like the following
wage married
. 1
1 1
1 1
3 0
4 0
or
wage married
. 1
. 1
1 1
3 0
4 0
Due to the random nature of this sample selection, we may have different
number of observations for the main equation each time a random sample is
selected.
Did I understand the bstrap procedure correctly? Is it applicable to a
two-step estimation?
Is this something that should be concerned about?
Thank you in advance for your comments and explanations.
Guanghui Li
*
* 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/
