Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Appropriate lags for Augmented Dickey-Fuller Test


From   "Joao Ricardo F. Lima" <jricardofl@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Appropriate lags for Augmented Dickey-Fuller Test
Date   Thu, 27 Aug 2009 16:37:22 -0300

Ihtesham,

what´s the null hyphothesis that you are testing?
Ho: lCPI exhibits a unit root.

With these dfgls test statistics you can´t reject H0 at 10%. Test
Statistic < Critical Value for lags 1-15.

ICPI isn´t stationary.

HTH,

Joao Lima

2009/8/27 Ihtesham Afzal <ihtesham_afzal@hotmail.co.uk>:
> Thank you again.
>
> If I do decide to go with the -dfgls- command, could you please help me interpret the results? i.e. is the time-series stationary?
> . dfgls lCPI
>
> DF-GLS for lCPI                                          Number of obs =   284
> Maxlag = 15 chosen by Schwert criterion
>
>               DF-GLS tau      1% Critical       5% Critical      10% Critical
>  [lags]     Test Statistic        Value             Value             Value
> ------------------------------------------------------------------------------
>    15           -0.614           -3.480            -2.815            -2.535
>    14           -0.659           -3.480            -2.823            -2.542
>    13           -0.787           -3.480            -2.830            -2.549
>    12           -1.235           -3.480            -2.838            -2.555
>    11           -0.351           -3.480            -2.845            -2.562
>    10           -0.285           -3.480            -2.851            -2.568
>    9            -0.223           -3.480            -2.858            -2.574
>    8            -0.327           -3.480            -2.865            -2.580
>    7            -0.381           -3.480            -2.871            -2.586
>    6            -0.457           -3.480            -2.877            -2.591
>    5            -0.254           -3.480            -2.883            -2.596
>    4            -0.164           -3.480            -2.888            -2.601
>    3            -0.084           -3.480            -2.894            -2.606
>    2            -0.233           -3.480            -2.899            -2.611
>    1            -0.206           -3.480            -2.903            -2.615
>
> Opt Lag (Ng-Perron seq t) = 13 with RMSE  .0030169
> Min SC   = -11.32856 at lag 13 with RMSE  .0030169
> Min MAIC = -11.51094 at lag 14 with RMSE   .003008
>
>
> Kind Regards
> Ihtesham
>
> ----------------------------------------
>> Date: Thu, 27 Aug 2009 16:31:36 +0100
>> Subject: Re: st: Appropriate lags for Augmented Dickey-Fuller Test
>> From: tirthankar.chakravarty@gmail.com
>> To: statalist@hsphsun2.harvard.edu
>>
>> -pperron- does not use lags to correct for serial correlation in the
>> Dickey-Fuller regressions and instead uses a Newey-West type robust
>> variance-covariance matrix to calulate the standard errors. However,
>> please see [TS] pperron documentation for cases when the
>> Phillips-Perron test is not applicable.
>>
>> If you are going to use -dfuller-, you might look at -varsoc- for
>> lag-length selection.
>>
>> You also do not specify a trend or drift terms in your -dfuller-. You
>> need to be sure that this is indeed the case, a very good preliminary
>> strategy is to visualise the data to see if there is a deterministic
>> trend in it.
>>
>> T
>>
>> 2009/8/27 Ihtesham Afzal :
>>>
>>> Thanks again for your reply.
>>> I think I will use the dfuller, lags(#) command as I think that is sufficient for the test I am lookinf to do
>>> Just one question though. I was made aware that I dont need to include a lags(#) term for the Phillips-Perron test does not need the lagged values.
>>> Is this correct?
>>> Thank you
>>> Regards
>>> Ihtesham
>>> ----------------------------------------
>>>> Date: Thu, 27 Aug 2009 13:58:57 +0100
>>>> Subject: Re: st: Appropriate lags for Augmented Dickey-Fuller Test
>>>> From: tirthankar.chakravarty@gmail.com
>>>> To: statalist@hsphsun2.harvard.edu
>>>>
>>>> Not really. -dfuller- provides what are called parametric augmented
>>>> Dickey-Fuller (ADF) tests, which require explicit specification of the
>>>> (linear) trend component. -dfgls- belongs to a class of efficient ADF
>>>> unit root tests, which involve running the ADF tests on
>>>> quasi-differenced series (wherby a trend is extracted by GLS
>>>> detrending) - leading, typically, to greater power properties.
>>>>
>>>> In either case, -dfuller- does not include a trend by default and
>>>> -dfgls- does. So as you are using them, they are not comparable. For
>>>> quick diagnostics on unit roots, I prefer -dfgls-.
>>>>
>>>> I can point you to the excellent exposition in Phillips and Xiao
>>>> (1998) & [TS] dfgls.
>>>>
>>>> --- References ---
>>>> @article{phillips1998primer,
>>>> title={{A Primer on Unit Root Testing}},
>>>> author={Phillips, P.C.B. and Xiao, Z.},
>>>> journal={Journal of Economic Surveys},
>>>> volume={12},
>>>> number={5},
>>>> pages={423--470},
>>>> year={1998},
>>>> publisher={Blackwell Publishers Ltd}
>>>> }
>>>>
>>>>
>>>>
>>>> 2009/8/27 Ihtesham Afzal :
>>>>> Hello, first of all, thanks for the reply.
>>>>> Here is the output. From this can I infer that the lags i should use is 14 (if I use the MAIC)and thus conduct the ADF test with 14 lags as I have done below?
>>>>> Is this the correct procedure.
>>>>> Kinds Regards.
>>>>> Ihtesham
>>>>>
>>>>>
>>>>> . dfgls lCPI
>>>>>
>>>>> DF-GLS for lCPI Number of obs = 284
>>>>> Maxlag = 15 chosen by Schwert criterion
>>>>>
>>>>> DF-GLS tau 1% Critical 5% Critical 10% Critical
>>>>> [lags] Test Statistic Value Value Value
>>>>> ------------------------------------------------------------------------------
>>>>> 15 -0.614 -3.480 -2.815 -2.535
>>>>> 14 -0.659 -3.480 -2.823 -2.542
>>>>> 13 -0.787 -3.480 -2.830 -2.549
>>>>> 12 -1.235 -3.480 -2.838 -2.555
>>>>> 11 -0.351 -3.480 -2.845 -2.562
>>>>> 10 -0.285 -3.480 -2.851 -2.568
>>>>> 9 -0.223 -3.480 -2.858 -2.574
>>>>> 8 -0.327 -3.480 -2.865 -2.580
>>>>> 7 -0.381 -3.480 -2.871 -2.586
>>>>> 6 -0.457 -3.480 -2.877 -2.591
>>>>> 5 -0.254 -3.480 -2.883 -2.596
>>>>> 4 -0.164 -3.480 -2.888 -2.601
>>>>> 3 -0.084 -3.480 -2.894 -2.606
>>>>> 2 -0.233 -3.480 -2.899 -2.611
>>>>> 1 -0.206 -3.480 -2.903 -2.615
>>>>>
>>>>> Opt Lag (Ng-Perron seq t) = 13 with RMSE .0030169
>>>>> Min SC = -11.32856 at lag 13 with RMSE .0030169
>>>>> Min MAIC = -11.51094 at lag 14 with RMSE .003008
>>>>>
>>>>>
>>>>> . dfuller lCPI, lags(14)
>>>>> Augmented Dickey-Fuller test for unit root Number of obs = 285
>>>>> ---------- Interpolated Dickey-Fuller ---------
>>>>> Test 1% Critical 5% Critical 10% Critical
>>>>> Statistic Value Value Value
>>>>> ------------------------------------------------------------------------------
>>>>> Z(t) -1.980 -3.457 -2.879 -2.570
>>>>> ------------------------------------------------------------------------------
>>>>> MacKinnon approximate p-value for Z(t) = 0.2955
>>>>>
>>>>>
>>>>> ----------------------------------------
>>>>>> Date: Thu, 27 Aug 2009 12:59:34 +0100
>>>>>> Subject: Re: st: Appropriate lags for Augmented Dickey-Fuller Test
>>>>>> From: tirthankar.chakravarty@gmail.com
>>>>>> To: statalist@hsphsun2.harvard.edu
>>>>>>
>>>>>> <>
>>>>>> -dfgls- reports three different criteria for lag selection:
>>>>>>
>>>>>> 1) Ng-Perron
>>>>>> 2) Schwarz
>>>>>> 3) Modified AIC
>>>>>>
>>>>>> and reports tests upto a max. lag determined by the Schwert criteria.
>>>>>>
>>>>>> T
>>>>>>
>>>>>> On Thu, Aug 27, 2009 at 12:39 PM, Ihtesham
>>>>>> Afzal wrote:
>>>>>>> Hello.
>>>>>>> Just a quick question.
>>>>>>> When undergoing the Augmented Dickey Fuller Test, how do I decide on how many lags to include for each series?
>>>>>>> Do I estimate the AR(p) model with different p-lag values and then find the one with the lowest AIC/BIC value?
>>>>>>>
>>>>>>> Kind Regards.
>>>>>>>
>>>>>>> Ihtesham
>>>>>>> _________________________________________________________________
>>>>>>>
>>>>>>> Upgrade to Internet Explorer 8 Optimised for MSN.
>>>>>>>
>>>>>>> http://extras.uk.msn.com/internet-explorer-8/?ocid=T010MSN07A0716U
>>>>>>> *
>>>>>>> * 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/
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> To every ω-consistent recursive class κ of formulae there correspond
>>>>>> recursive class signs r, such that neither v Gen r nor Neg(v Gen r)
>>>>>> belongs to Flg(κ) (where v is the free variable of r).
>>>>>>
>>>>>> *
>>>>>> * 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/
>>>>> _________________________________________________________________
>>>>> Celebrate a decade of Messenger with free winks, emoticons, display pics, and more.
>>>>> http://clk.atdmt.com/UKM/go/157562755/direct/01/
>>>>> *
>>>>> * 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/
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> To every ω-consistent recursive class κ of formulae there correspond
>>>> recursive class signs r, such that neither v Gen r nor Neg(v Gen r)
>>>> belongs to Flg(κ) (where v is the free variable of r).
>>>>
>>>> *
>>>> * 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/
>>> _________________________________________________________________
>>> Windows Live Messenger: Happy 10-Year Anniversary-get free winks and emoticons.
>>> http://clk.atdmt.com/UKM/go/157562755/direct/01/
>>> *
>>> * 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/
>>>
>>
>>
>>
>> --
>> To every ω-consistent recursive class κ of formulae there correspond
>> recursive class signs r, such that neither v Gen r nor Neg(v Gen r)
>> belongs to Flg(κ) (where v is the free variable of r).
>>
>> *
>> * 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/
> _________________________________________________________________
> Windows Live Messenger: Thanks for 10 great years-enjoy free winks and emoticons.
> http://clk.atdmt.com/UKM/go/157562755/direct/01/
> *
> *   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/
>



-- 
----------------------------------------
Joao Ricardo Lima, D.Sc.
Professor
UFPB-CCA-DCFS
Fone: +5538387264913
Skype: joao_ricardo_lima
----------------------------------------

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index