Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: the newey2, number of lags and by function

 From yilmag@rpi.edu To statalist@hsphsun2.harvard.edu Subject st: the newey2, number of lags and by function Date Sun, 13 Nov 2011 0:51:45 -0500

```Dear Listers,

I am running the 4-factor model to detect the abnormal returns of two
portfolios using monthly returns data. My panelvar is the gvkey (firm id)
and the timevar is from 2007m4 to 2008m3. Here is what I do:

tsset gvkey time

panel variable:  gvkey (strongly balanced)

time variable:  time, 2007m4 to 2008m3

delta:  1 month

//Carhart 4-Factor Model using equally weighted returns for each portfolio

newey2 ewxsret xsewretd smb hml umd if inlist(group,0), lag(3) force

newey2 ewxsret xsewretd smb hml umd if inlist(group,1), lag(3) force

//Carhart 4-Factor Model using value weighted returns for each portfolio

newey2 vwxsret xsvwretd smb hml umd if inlist(group,0), lag(3) force

newey2 vwxsret xsvwretd smb hml umd if inlist(group,1), lag(3) force

Basically the newey2 runs the OLS for each firm-month and averages across
time for a given group. Here are my questions:

1)In order to determine the optimal #of lags (I just used 3 but I don't
have a justification), I try to use the post-estimation "varsoc" command
but I get "either specify a varlist or the estimates() from either var or
svar". When I specify the varlist as ewxsret, I get "repeated time values
in sample" because ewxsret (equally weighted excess returns) is the same
for all firms in a given portfolio (group) in a given month. How should I
proceed with this?

2)How can I compare the constant terms from the above regressions? In other
words, for equal weighted and value weighted settings, I want to see if the
abnormal returns to two portfolios are statistically different from each
other. I tried to use the "suest" command but it says that the models are
estimated with a "nonstandard vce (Newey-West)". I want to stick with
Newey-West because returns are heteroskedastic and autocorrelated.

3)Is there a way to run the 4-factor model for both portfolios(groups) in
one regression using newey2? I tried by/bys but it says that "newey2 may
not be combined with by"

Thanks everyone in advance for the help.

Gokhan

*
*   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/
```