I think I sended it to R-Sig-Finance ([email protected])...

Well... I analyzed the series with the Augmented Dickey Fuller test. I used the 
procedure from Dolado, Jenkinson and Rivero (1990) to identify the unit root.

Some idea:
Can I model the non-stationary time series with the Augmented Dickey Fuller 
model and use the intercept and the standard error of the intercept as proxies 
for the mean of the series? That would be something like a conditional mean 
"given the series is stationary"...
Can I calculate then robust standard errors with the sandwich-package of that 
Augmented Dickey Fuller-Intercept? -> vcovHC(ADFmodel, type="HC0") or 
NeweyWest(ADFmodel)


Andy






--- [email protected] <[email protected]> schrieb am So, 16.5.2010:

Von: [email protected] <[email protected]>
Betreff: Re: [R-SIG-Finance] Robust standard error for a time series mean.
An: [email protected]
Datum: Sonntag, 16. Mai, 2010 01:27 Uhr

Hi: i'm not sure what you mean by "non-stationary" ( there are many versions of 
non-stationarity ) but, if you mean non-constant mean , then I don't think the 
standard error has any meaning and you need to get stationary first.  But , if 
it's returns, I'm not sure
how you would do that because I thought returns were generally stationary ( 
!~zero mean ).

The book by Jan Beran, "statistics for long memory processes"  has some  
material on calculating the standard error for the mean of a time series when 
the values are not independent but it may be similar to the Newey West 
estimator. I'm not sure abut that but the book may be worth looking at. 
Hopefully someone will else will say more. Another thing: it may be better to 
send that question to R-Sig-Finance because you may get more responses. good 
luck.









hi: I'
On May 15, 2010, Andreas Klein <[email protected]> wrote: Hello.

I have a non-stationary time series of returns and I would like to calculate 
the standard error for the mean of that series.

When I use the White-estimator or assume constant variance I got the same 
results. When I use the Newey-West-estimator, which also cares about 
autocorrelation, the standard errors increase a lot.

1. What is the right estimator?
2. Is the Newey West estimator strongly affected by the non-stationarity?


Does anyone have any literature source or a link to a paper or something like 
that? 


I hope someone can help me.


Thank you in advance.


Sincerely,
Andy.



_______________________________________________
[email protected] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should 
go.




        [[alternative HTML version deleted]]

_______________________________________________
[email protected] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should 
go.

Reply via email to