Which one is better?

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Which one is better?

KAUSHIK BHATTACHARJEE
I have some time series .more than 1000 observations each ..I want to test if  they contain unit roots...
I am performing ADF as well as  KPSS tests...I am getting contraditory results in 7 out of 10 series....KPSS test is rejecting stationarity where as ADF test is not....
Which one is more reliable test...any idea/reference?
 
Kaushik Bhattacharjee



     
        [[alternative HTML version deleted]]


_______________________________________________
[hidden email] 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 | Threaded
Open this post in threaded view
|

Re: Which one is better?

Matthieu Stigler
KAUSHIK BHATTACHARJEE a écrit :
> I have some time series .more than 1000 observations each ..I want to test if  they contain unit roots...
> I am performing ADF as well as  KPSS tests...I am getting contraditory results in 7 out of 10 series....KPSS test is rejecting stationarity where as ADF test is not....
> Which one is more reliable test...any idea/reference?
>  
>  
any reference? Yes, the paper of KPSS! It is not too technical and in
the empirical part they discuss this issue of contradictory results with
unit roots tests (if I remember well there are five cases from the
Nelson Plosser data were KPSS is different than ADF, so that's a known
issue:-).

I would rather base on a unit root test (btw rather ERS than ADF) as the
KPSS test is rather considered as a "confirmatory test". Base on unit
root tests especially when both tests are rejected, as unit root tests
are known to have low power and rejection is then a pretty "big sign".
But note that their size is not so good neither, while KPSS can have
really big size distortions, try yourself:

library(urca)

simul.cval<-function(ar, n=100){
  series<-arima.sim(model=list(ar=ar), n=n)
  summary(ur.kpss(series, use.lag=1))@teststat
}

mc<-replicate(500, simul.cval(ar=0.2))
mean(mc>  0.463)

mc<-replicate(500, simul.cval(ar=0.9))
mean(mc>  0.463)

Hope this helps

Mat

> Kaushik Bhattacharjee
>
>
>
>      
> [[alternative HTML version deleted]]
>
>  
> ------------------------------------------------------------------------
>
> _______________________________________________
> [hidden email] 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.

_______________________________________________
[hidden email] 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 | Threaded
Open this post in threaded view
|

Re: Which one is better?

Patrick Burns-2
I suspect this is asking the wrong question
of the data.  I don't know what the right
questions are, but they should be in the
direction of where ever you want to go.

mat wrote:

> KAUSHIK BHATTACHARJEE a écrit :
>> I have some time series .more than 1000 observations each ..I want to
>> test if  they contain unit roots...
>> I am performing ADF as well as  KPSS tests...I am getting contraditory
>> results in 7 out of 10 series....KPSS test is rejecting stationarity
>> where as ADF test is not....
>> Which one is more reliable test...any idea/reference?
>>  
>>  
> any reference? Yes, the paper of KPSS! It is not too technical and in
> the empirical part they discuss this issue of contradictory results with
> unit roots tests (if I remember well there are five cases from the
> Nelson Plosser data were KPSS is different than ADF, so that's a known
> issue:-).
>
> I would rather base on a unit root test (btw rather ERS than ADF) as the
> KPSS test is rather considered as a "confirmatory test". Base on unit
> root tests especially when both tests are rejected, as unit root tests
> are known to have low power and rejection is then a pretty "big sign".
> But note that their size is not so good neither, while KPSS can have
> really big size distortions, try yourself:
>
> library(urca)
>
> simul.cval<-function(ar, n=100){
>  series<-arima.sim(model=list(ar=ar), n=n)
>  summary(ur.kpss(series, use.lag=1))@teststat
> }
>
> mc<-replicate(500, simul.cval(ar=0.2))
> mean(mc>  0.463)
>
> mc<-replicate(500, simul.cval(ar=0.9))
> mean(mc>  0.463)
>
> Hope this helps
>
> Mat
>> Kaushik Bhattacharjee
>>
>>
>>
>>           [[alternative HTML version deleted]]
>>
>>  
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> [hidden email] 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.
>
> _______________________________________________
> [hidden email] 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.
>
>

_______________________________________________
[hidden email] 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 | Threaded
Open this post in threaded view
|

Re: Which one is better?

John C. Frain
In reply to this post by KAUSHIK BHATTACHARJEE
Without knowing something about the nature of your data and the
purpose of your analysis this is a hard question to answer.  Do you
have an underlying theory agout the behaviour of your series?   If
that theory implies a unit root then and ADF type test (probably an
ERS test, as recommended by Mat)  is an appropriate choice.  If theory
says that the series is stationary then the KPSS test is an
appropriate test of the null of stationarity.  There are, of course,
non-stationary series that do not have a unit root.  It is nice if ADF
and KPSS tests lead to the same conclusion but there is no
contradiction if this does not happen.

Maddala and Kim (1998), Unit Roots Cointegration and Structural
Change, Cambridge University Press, is a good suvey of the theory of
unit roots etc.

Leybourne and McCabe (1989) shows how certain revisions to the KPSS
test can make the results of the ADF and KPSS test more consistent.

John




2010/1/1 KAUSHIK BHATTACHARJEE <[hidden email]>:

> I have some time series .more than 1000 observations each ..I want to test if  they contain unit roots...
> I am performing ADF as well as  KPSS tests...I am getting contraditory results in 7 out of 10 series....KPSS test is rejecting stationarity where as ADF test is not....
> Which one is more reliable test...any idea/reference?
>
> Kaushik Bhattacharjee
>
>
>
>
>        [[alternative HTML version deleted]]
>
>
> _______________________________________________
> [hidden email] 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.
>



--
John C Frain, Ph.D.
Trinity College Dublin
Dublin 2
Ireland
www.tcd.ie/Economics/staff/frainj/home.htm
mailto:[hidden email]
mailto:[hidden email]

_______________________________________________
[hidden email] 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.