Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "Dimitriy V. Masterov" <dvmaster@gmail.com> |
To | Statalist <statalist@hsphsun2.harvard.edu> |
Subject | st: robust hypothesis test for totals |
Date | Wed, 13 Mar 2013 10:23:24 -0700 |
I would like to estimate a multiplicative treatment effect for a total (rather than a mean) with a confidence interval. The outcome is very left-skewed, with lots of zeros and a kangaroo-like tail. Alternatively, I would like to conduct a hypothesis that the effect is 1. Here's an example of what I am thinking of: sysuse auto bs ratio = _b[Domestic]/_b[Foreign], reps(100): total price, over(foreign) Is bootstrapping the best approach or is there something better? Can this or some other procedure be generalized to more than two groups? * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/