st: Sensitivity analysis with MHBOUNDS

[Shikha Sinha <shikha.sinha414@gmail.com>]

Dear Statalist,

I am using  -mhbounds to test the sensitivty of ATT to "hidden bias".
The treatment is "access to electricity" and outcomes are health
variables. I am new to this, amd not sure if I should look at Q_mh+ or
Q_mh-.

A paper by Becker and Caliendo (2007) in Stata journal states that -
in case of positive selection , Q_mh+ shold be examined and for
negative selection, Q_mh- shoul dbe examined.

What exactly are the positive and negative selection in my example? I
am examining the effect on electricity on health outcomes. Results
show that access to electricity reduced health risk caused from smoke.
Does negative selection means that "households who are most likely to
have access to electricity may also have lower smoke-induced health
risk.?
It may also be true that "households with electricity also have higher
probability of health risk"?

Please elaborate on this positive and negative selection in the
-mhbounds context. Is there any test to check the type of selection,
how do we decide about the type of selection?

Thanks,
Shikha
*
*   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/