Dear all,
Trying to test for heteroskedasticity I tried several test from the
car package respectively lmtest. Now that they produce rather
different results i am somewhat clueless how to deal with it.
Here is what I did:
1. I plotted fitted.values vs residuals and somewhat intuitively
believe, it isn't really increasing...
2. further I ran the following tests
bptest (studentized and non-studentized), gqtest, ncv.test with the
following results:
ncv:
Non-constant Variance Score Test
Variance formula: ~ fitted.values
Chisquare = 13.87429 Df = 1 p = 0.00194580
Goldfeld-Quandt test
data: reg
GQ = 1.7092, df1 = 327, df2 = 327, p-value = 7.93e-07
studentized Breusch-Pagan test
data: reg
BP = 15.8291, df = 23, p-value = 0.92
Breusch-Pagan test
data: reg
BP = 377.5604, df = 23, p-value < 2.8e-18
bptest and gq.test sport pretty straight forward examples saying the
H0 = homoskedasticity. The ncv.test clarifies the same in its
description. Thus the studentized bptest appears to be the only one to
support my first intuition from the graphical solution. I am really
disturbed what to do know and how to interpret my results... Can
someone lead the way ?
thx in advance
matt
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.