Dear All,
I am using the heckman correction for selectivity bias. The type of selection bias I am adjusting for is sometimes called heterogeneity bias.
The specific issue is as follows. I intend looking at the impact of receiving a poverty alleviation grant s, on child weight for height w.
If I were running an OLS the model would take the form:
W=f(s Z) where s is binary and Z is a set of control variables. If the coefficient on the dummy s were significant I would have concluded, simply put, that the grant had an impact on child weight.
But there is selection bias because receiving the poverty alleviation grant may depend on the weight of children. I attempt correcting for this as follows:
heckman( w Z), select (s=Z X) where X influences s but not w.
1) Is this correct?
2) How do I now assess whether the grant s has a significant impact on w given that s is no longer in the substantial equation?
3) the results say that 483 observations are censored and 250 are not. Does this mean that the selection equation is based on 250?
4) why does the twostep option give a different result?
5) If s were not 0 or 1, but instead a positive variable indicating for how long you've received the grant, can I still use the above heckman command?
I deeply appreciate your advise and support. Thanks in advance.
bets regards,
sally.
*
* 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/