[R] multiple break in univariate series

[Achim Zeileis]

On Sat, 28 Mar 2015, Bert Gunter wrote:

> Once you have looked at the data and chosen change points to test
> based on the data, the tests for change points are invalid (unless you
> make appropriate adjustments for post hoc tests).
>
> And no, I am not making this up. Consult any competent statistician.

Yes, you must not select a breakpoint by eyeballing a time series and then 
conduct a structural break test for this given breakpoint as if it were 
exogenously given. This will be far too liberal.

But most practitioners would think it is still ok to conduct a structural 
break test with _unknown_ breakpoint. Ideally, the hypothesis to be tested 
and the significance level were formulated prior to the collection of the 
data, though.

To add to Uwe's comments: maxstat_test() in package "coin" would be a 
non-parametric permutation test. sctest() in package "strucchange" 
provides a wide range of tests for general parametric models, especially 
linear regressions.

Best,
Z

> Cheers,
> Bert
>
>
>
>
>
> Bert Gunter
> Genentech Nonclinical Biostatistics
> (650) 467-7374
>
> "Data is not information. Information is not knowledge. And knowledge
> is certainly not wisdom."
> Clifford Stoll
>
>
>
>
> On Sat, Mar 28, 2015 at 5:52 PM, Uwe Ligges
> <ligges at statistik.tu-dortmund.de> wrote:
>>
>>
>> On 29.03.2015 00:09, Temel ?spanyolca wrote:
>>>
>>>
>>>     DR. UWE LIGGES
>>>
>>>
>>> I have sent turkish real Gdp data (1998-2013) in annex.
>>> Turkey has lived two crises in this period, in 2001 and 2008. You can
>>> see in data.
>>> My problem is to indicate these dates any statistic test as a structural
>>> change or point change.
>>
>>
>> Have you tried CUSUM? Or some permutations test? You do not have muh data
>> ....
>>
>> Best,
>> Uwe Ligges
>>
>>
>>
>>>
>>> Sincerely
>>> Engin
>>>
>>> 2015-03-28 23:04 GMT+01:00 Uwe Ligges <ligges at statistik.tu-dortmund.de
>>> <mailto:ligges at statistik.tu-dortmund.de>>:
>>>
>>>             -------- Forwarded Message --------
>>>             Subject: [R] multiple break in univariate series
>>>             Date: Sat, 28 Mar 2015 00:41:35 +0100
>>>             From: Temel ?spanyolca <ispanyolcom at gmail.com
>>>             <mailto:ispanyolcom at gmail.com>>
>>>             To: R-help at r-project.org <mailto:R-help at r-project.org>
>>>
>>>             Hello
>>>             Any one knows multiple break test for univariate series ?
>>>
>>>
>>>
>>>     Which kind of breaks? shift?
>>>
>>>     You may want to look for CUSUM or MOSUM tests or even permutation
>>> tests.
>>>
>>>     Packages: strucchange, changepoint
>>>
>>>     http://cran.r-project.org/web/__packages/changepoint/index.__html
>>>     <http://cran.r-project.org/web/packages/changepoint/index.html>
>>>
>>>     http://cran.r-project.org/web/__packages/strucchange/index.__html
>>>     <http://cran.r-project.org/web/packages/strucchange/index.html>
>>>
>>>     Best,
>>>     Uwe Ligges
>>>
>>>
>>>
>>>
>>>             --
>>>             *Thanks*
>>>             Engin YILMAZ
>>>
>>>                       [[alternative HTML version deleted]]
>>>
>>>             ________________________________________________
>>>             R-help at r-project.org <mailto:R-help at r-project.org> mailing
>>>             list -- To UNSUBSCRIBE and more, see
>>>             https://stat.ethz.ch/mailman/__listinfo/r-help
>>>             <https://stat.ethz.ch/mailman/listinfo/r-help>
>>>             PLEASE do read the posting guide
>>>             http://www.R-project.org/__posting-guide.html
>>>             <http://www.R-project.org/posting-guide.html>
>>>             and provide commented, minimal, self-contained, reproducible
>>>             code.
>>>
>>>
>>>
>>>
>>>
>>> --
>>> *Sayg?lar?mla*
>>> Engin YILMAZ
>>
>>
>> ______________________________________________
>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> 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.
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.