On Fri, Jan 18, 2013 at 1:34 PM, Nahla Betelmal wrote:
> Also, the
> normality test show rejection of normality both before and after I
> winsorized the data
That is not surprising. Based on the simulation I gave you yesterday
you can see that if you had a 1000 observations (rather than 346) and
the data really came from a Gaussian distribution you would still have
a 93% chance of rejecting that true hypothesis at the 5% level!
Rejecting the null hypothesis based on these tests is thus completely
and utterly and totally uninformative. So, don't use these tests in
such small samples.
-- Maarten
---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
Germany
http://www.maartenbuis.nl
---------------------------------
*
* 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/
