|
|  |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: estat hettest: Breusch-Pagan Test
suggested by Mark Schaffer) note that the White/Koenker
heteroskedasticity test does not require the assumption of normality
that the Breusch-Pagan/Godfrey/Cook-Weisberg test (implemented in
-estat hettest-) does.
This fact appears to contribute to the observed differences between
the two test statistics in my prior messages:
use "http://fmwww.bc.edu/ec-p/data/wooldridge/HPRICE1";, clear
reg lprice llotsize lsqrft bdrms
predict uhat, resid
sktest lprice uhat
swilk lprice uhat
Skewness/Kurtosis tests for Normality
------- joint ------
Variable | Pr(Skewness) Pr(Kurtosis) adj chi2(2) Prob>chi2
-------------+-------------------------------------------------------
lprice | 0.020 0.014 9.78 0.0075
uhat | 0.444 0.001 10.25 0.0059
Shapiro-Wilk W test for normal data
Variable | Obs W V z Prob>z
-------------+-------------------------------------------------
lprice | 88 0.95205 3.560 2.797 0.00258
uhat | 88 0.95184 3.576 2.807 0.00250
I'm still somewhat surprised by the sharp difference between the two
statistics in this example, but I appreciate the information provided
by Scott, Mark and David to clarify this matter. Based on these
results, it seems that the correct inference is provided by the
White/Koenker test, which does not reject the null of homoskedasticity.
-- Mike
*
* 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/
