Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: ZANDREWS


From   "Joao Ricardo F. Lima" <jricardofl@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: ZANDREWS
Date   Sun, 14 Sep 2008 21:00:31 -0300

Richard,

I think that you can do that. However, isnīt the default procedure.
When I am working with time series, I seek to know if something
happened that it caused the shift level. Example: I'm from Brazil and
1999 is important because of the depreciation of the exchange tax.
2002 is important too, because the elections for president, etc.

JL



2008/9/14 Richard Harvey <richardharvey2008@googlemail.com>:
> Hi,
>
> Thanks for the link. I am not sure Chow test is the way forward for
> me... primarily because it assumes you know when the break occurs. In
> the example on that page we know exactly at which point one group ends
> and the other starts.  What one wants is to identify the breaks
> without assuming that one knows the exact break points. However, could
> one apply the chow test to all the time points between a range and see
> which turns out to be significant?
>
>
>
>
>
> 2008/9/14 Joao Ricardo F. Lima <jricardofl@gmail.com>:
>> Richard,
>>
>> to do the Chow test, I always see this FAQ:
>>
>> http://www.stata.com/support/faqs/stat/chow.html
>>
>> HTH,
>>
>> Joao Lima
>>
>> 2008/9/14 Richard Harvey <richardharvey2008@googlemail.com>:
>>> Thanks robert, maarten and joao,
>>>
>>> I am not entirely sure I understand how prof. yafee could identify the
>>> two breaks he pointed out. I couldnt understand what the statement "If
>>> we use a 3.5 sd measure of a critical value for a standard of a level
>>> shift" ..
>>>
>>> Thanks for the clemao2  suggestion, I will  try it out.
>>>
>>> But i have an issue with the chow test. As I under stand it uses OLS
>>> regression using points before and after the break point. The
>>> relaibaility of this test would be highly dependent on when the break
>>> happens, is it not?  there must be enough number of time series points
>>> to run the regressions reliably.
>>>
>>> There is an ado which does the chow test in stata, I was looking at it
>>> when I got your mail. But the ado by Sean Becketti - STB:  sts7
>>> (STB-17) has errors. When I run it, I get the following
>>>
>>> program error:  code follows on the same line as open brace
>>> r(198);
>>>
>>> I tried to fix the ado, but now it dosent give me any results,
>>> nothing. Set trace on reveals nothing.
>>>
>>> thanks
>>> rich
>>>
>>>
>>>
>>>
>>>
>>> On 9/14/08, Joao Ricardo F. Lima <jricardofl@gmail.com> wrote:
>>>> Dear Richard,
>>>>
>>>> if you are looking for structural breaks in the data, why donīt you
>>>> use Chow Test?
>>>>
>>>> As I understand, Zandrews is a unit root test allowing for structural
>>>> change. According to Professor Robert, there are two significant level
>>>> shifts in the series. Zandrews test is for only one break. With 2
>>>> breaks you can use clemao2 or clemio2 -findit clemao_io-.
>>>>
>>>> See Kitīs paper "Stata: the languague of choice for time series
>>>> analysis?" to understand these unit root tests. You can download it
>>>> free:
>>>>
>>>> http://www.stata-journal.com/article.html?article=st0080
>>>>
>>>> Hope this helps,
>>>>
>>>> Joao Lima
>>>>
>>>> 2008/9/13 Richard Harvey <richardharvey2008@googlemail.com>:
>>>>> hello statalisters,
>>>>>
>>>>> I have the following time series data
>>>>>
>>>>> Months    MNval
>>>>> -36     -246404.3
>>>>> -35     -271136.9
>>>>> -34     15321.21
>>>>> -33     -266850.2
>>>>> -32     -208945.8
>>>>> -31     -223758.4
>>>>> -30     -152778.7
>>>>> -29     -65240.22
>>>>> -28     -481321.3
>>>>> -27     -190169.6
>>>>> -26     -112362.4
>>>>> -25     -467139.1
>>>>> -24     -194511.7
>>>>> -23     -444933.2
>>>>> -22     -142801.8
>>>>> -21     -108734.8
>>>>> -20     -217134.7
>>>>> -19     -285050.4
>>>>> -18     -59074.71
>>>>> -17     -22606.87
>>>>> -16     -124248.8
>>>>> -15     -190182.6
>>>>> -14     -142432.7
>>>>> -13     -137051
>>>>> -12     55313.91
>>>>> -11     -33965.3
>>>>> -10     -274899.1
>>>>> -9      25388.8
>>>>> -8      -57483.93
>>>>> -7      -39486.45
>>>>> -6      -54602.45
>>>>> -5      -239732.4
>>>>> -4      -294504.7
>>>>> -3      -334388.7
>>>>> -2      -287347.5
>>>>> -1      -87598.81
>>>>> 0       -580743.6
>>>>> 1       -160395.5
>>>>> 2       -575569.9
>>>>> 3       -255174.5
>>>>> 4       -126203.1
>>>>> 5       -375763.7
>>>>> 6       -32561.49
>>>>> 7       -87045.1
>>>>> 8       -91822.47
>>>>> 9       -31134.7
>>>>> 10      -102519
>>>>> 11      -84379.84
>>>>> 12      -266297.5
>>>>> 13      -80520.79
>>>>> 14      -48140.91
>>>>> 15      27969.55
>>>>> 16      -38107.77
>>>>> 17      -127254
>>>>> 18      -171679.9
>>>>> 19      -25015.38
>>>>> 20      -92405.42
>>>>> 21      75767.31
>>>>> 22      -45497.32
>>>>> 23      -188625.2
>>>>> 24      30972.86
>>>>> 25      -4018.882
>>>>> 26      -32145.2
>>>>> 27      -80913.73
>>>>> 28      -24621.37
>>>>> 29      -61688.87
>>>>> 30      -123890.4
>>>>> 31      -45585.37
>>>>> 32      -132729.1
>>>>> 33      -106762.2
>>>>> 34      -143616.2
>>>>> 35      -215439.4
>>>>> 36      -133058.6
>>>>>
>>>>> I would like to see if and when there are any structural breaks in the
>>>>> data.
>>>>>
>>>>> I came across Kit's zandrews prog. -ssc d zandrews-  is this the
>>>>> command to use? the graph option on this looks very useful.
>>>>> But how does one interpret the break point t-stats? When i use the
>>>>> command with the above data.  I do not find any point above the
>>>>> (critical 5%: -4.42).
>>>>>  so does this mean  that the series does not have any significant
>>>>> breakpoints ?
>>>>>
>>>>> thanks very much for you time
>>>>> rich
>>>>> *
>>>>> *   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
>>>> Professor
>>>> UFPB-CCA-DCFS
>>>> +553138923914
>>>> -------------------------------
>>>>
>>>> *
>>>> *   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/
>>>>
>>>
>>> *
>>> *   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
>> Professor
>> UFPB-CCA-DCFS
>> +553138923914
>> -------------------------------
>>
>> *
>> *   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/
>>
>
>
>
> --
> thanks for your time
> rich
>
> *
> *   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
Professor
UFPB-CCA-DCFS
+553138923914
-------------------------------

*
*   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