Time series temporal disaggregation

[Axel Leroix]

 
Hi,
This is a newbie question.
I would to be able to convert annual time series of flow data into quarterly data. I wonder if there is any existing R-function which permits to do it? In what package ?
 
I the archive, i found that some poeple speak about "tempDis" package for performing time series temporal disaggregation, but when I try to download it I can not found it in the list of proposed packages.
 
Tank you in advance for your help.



     
        [[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.

[Brian G. Peterson]

Axel Leroix wrote:
> Hi,
> This is a newbie question.
> I would to be able to convert annual time series of flow data into quarterly data. I wonder if there is any existing R-function which permits to do it? In what package ?
>  
> I the archive, i found that some poeple speak about "tempDis" package for performing time series temporal disaggregation, but when I try to download it I can not found it in the list of proposed packages.
>  
> Thank you in advance for your help.
>  
Well, as discussed multiple times on this list, going from annual (or
any lower frequency) data to quarterly (or any higher frequency) data is
questionable at best.  Think data snooping or look-ahead bias in your
modeling.

Going the other direction, from say daily (or any higher frequency)  to
monthly (or any lower frequency) , is easily accomplished with to.period
for price/value data or Return.cumulative for returns data.

If you really do want to go in the black-magic direction of going from
annual to quarterly, first make sure that the "annual" data was not
first reported as monthly data or quarterly data (this is true for
almost all macroeconomic series) and then go back to the source data at
a higher frequency.

If even this is not possible, and you insist on the highly dubious
practice of taking an annual number and turning it into four quarterly
numbers, see the various na handling methods provided by the zoo
package, most likely na.approx or na.spline.

Regards,

     - Brian

--
Brian G. Peterson
http://braverock.com/brian/
Ph: 773-459-4973
IM: bgpbraverock

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only.
-- If you want to post, subscribe first.

[John C. Frain]

Several Official Bodies (Central Banks Eurostat etc) use Chow-Lin
interpolation to derive  quarterly data from annual or monthly from
quarterly.  It may be the case that the higher frequency data have
only recently been produced.  I don't know if anyone has produced any
R routines but if you google "Chow-Lin interpolation" you will
probably find implementations in Matlab or Gauss that should transfer
easily to R.  My implementation in RATS can  be accessed at
ideas.repec.org/p/cbi/wpaper/2-rt-04.html.  This contains an
explanation of the methodology and references to the original papers.

The research section of the Eurostat website also contained some
relevant material.

Best Regards

John

2009/10/30 Brian G. Peterson <[hidden email]>:

> Axel Leroix wrote:
>>
>> Hi,
>> This is a newbie question. I would to be able to convert annual time
>> series of flow data into quarterly data. I wonder if there is any existing
>> R-function which permits to do it? In what package ?
>>  I the archive, i found that some poeple speak about "tempDis" package for
>> performing time series temporal disaggregation, but when I try to download
>> it I can not found it in the list of proposed packages.  Thank you in
>> advance for your help.
>>
>
> Well, as discussed multiple times on this list, going from annual (or any
> lower frequency) data to quarterly (or any higher frequency) data is
> questionable at best.  Think data snooping or look-ahead bias in your
> modeling.
>
> Going the other direction, from say daily (or any higher frequency)  to
> monthly (or any lower frequency) , is easily accomplished with to.period for
> price/value data or Return.cumulative for returns data.
>
> If you really do want to go in the black-magic direction of going from
> annual to quarterly, first make sure that the "annual" data was not first
> reported as monthly data or quarterly data (this is true for almost all
> macroeconomic series) and then go back to the source data at a higher
> frequency.
>
> If even this is not possible, and you insist on the highly dubious practice
> of taking an annual number and turning it into four quarterly numbers, see
> the various na handling methods provided by the zoo package, most likely
> na.approx or na.spline.
>
> Regards,
>
>    - Brian
>
> --
> Brian G. Peterson
> http://braverock.com/brian/
> Ph: 773-459-4973
> IM: bgpbraverock
>
> _______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> -- Subscriber-posting only.
> -- If you want to post, subscribe first.
>
>

--
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.

[Brian G. Peterson]

John,

Thanks for the pointers.  I'll certainly research this method.

It appears on first reading that the Chow-Lin method requires
multivariate indicator series to attempt to remove biases potentially
introduced by the higher frequency indicators.  Do I understand this
correctly?

If so, is the resulting series truly "unbiased", or are the biases
introduced by construction via the indicator, and then minimized by the
distribution of the residual?

Also, you would not be able to decompose a univariate series of annual
numbers, as I thought was proposed by the original poster in this
thread, without identifying indicators.  Correct?

Thanks again,

  - Brian

John Frain wrote:

> Several Official Bodies (Central Banks Eurostat etc) use Chow-Lin
> interpolation to derive  quarterly data from annual or monthly from
> quarterly.  It may be the case that the higher frequency data have
> only recently been produced.  I don't know if anyone has produced any
> R routines but if you google "Chow-Lin interpolation" you will
> probably find implementations in Matlab or Gauss that should transfer
> easily to R.  My implementation in RATS can  be accessed at
> ideas.repec.org/p/cbi/wpaper/2-rt-04.html.  This contains an
> explanation of the methodology and references to the original papers.
>
> The research section of the Eurostat website also contained some
> relevant material.
>
> Best Regards
>
> John
>
> 2009/10/30 Brian G. Peterson <[hidden email]>:
>  
>> Axel Leroix wrote:
>>    
>>> Hi,
>>> This is a newbie question. I would to be able to convert annual time
>>> series of flow data into quarterly data. I wonder if there is any existing
>>> R-function which permits to do it? In what package ?
>>>  I the archive, i found that some poeple speak about "tempDis" package for
>>> performing time series temporal disaggregation, but when I try to download
>>> it I can not found it in the list of proposed packages.  Thank you in
>>> advance for your help.
>>>
>>>      
>> Well, as discussed multiple times on this list, going from annual (or any
>> lower frequency) data to quarterly (or any higher frequency) data is
>> questionable at best.  Think data snooping or look-ahead bias in your
>> modeling.
>>
>> Going the other direction, from say daily (or any higher frequency)  to
>> monthly (or any lower frequency) , is easily accomplished with to.period for
>> price/value data or Return.cumulative for returns data.
>>
>> If you really do want to go in the black-magic direction of going from
>> annual to quarterly, first make sure that the "annual" data was not first
>> reported as monthly data or quarterly data (this is true for almost all
>> macroeconomic series) and then go back to the source data at a higher
>> frequency.
>>
>> If even this is not possible, and you insist on the highly dubious practice
>> of taking an annual number and turning it into four quarterly numbers, see
>> the various na handling methods provided by the zoo package, most likely
>> na.approx or na.spline.
>>
>> Regards,
>>
>>    - Brian
>>
>> --
>> Brian G. Peterson
>> http://braverock.com/brian/
>> Ph: 773-459-4973
>> IM: bgpbraverock
>>
>> _______________________________________________
>> [hidden email] mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
>> -- Subscriber-posting only.
>> -- If you want to post, subscribe first.
>>
>>
>>    
>
>
>
>

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only.
-- If you want to post, subscribe first.

[John C. Frain]

Brian

Sorry for not answering your questions sooner.  I appologise if this reply is a bit off topic for this mailing list.

I have attached two pdf's which I hope explain in some way how we used the Chow-Lin interpoalation/distribution  methods.  In the Central Bank of Ireland we were interested  in modelling various aspects of the Irish Economy but quarterly national accounts were not available in Ireland until the late 90's.  In the early 80's I wrote  Chow-Lin routines first in TROLL and then in Gauss and calculated several sets of quarterly national accounts which were used in modelling various aspects of the economy.  

Manual.pdf describes a RATS program used to produce a set of national accounts for the Irish macro model component of the ECB systrm of macro-models.  The implementation assumes a certain relationship between the annual variable and the quarterly indicators.  If the model is valid then the estimates are unbiased.  As the model is probably not valid some bias certainly exists.  However analysis using the derived data has generally produced reasonable useful results.

The methodology used differs from the original Chow-Lin methodology.  Given the assumed model between the unobserved quarterly model and the indicators one can calculate the distribution of the annual data and estimate the parameters using maximum likelihood.

Manual2.pdf describes an extension of the methodology where quarterly data are available for some of the period and the likelhodd estimation is based on the distribution of the quarterly data where available and on the annual observations otherwise.  I think that tis method has been used in the Central Bank but I do not know the extent of that use as I retired from the Central Bank about 5 years ago.

The method can be used to decompose a series by using a constant and/or trend as indicators.  In most cases of interest there is some indicator.

In some cases one may set up some form of penalty function to be minimised and assume that the quarterly series follows some form of time series.

A long time ago we also experimented a little with Kalman Filters with limited sucess. These methods might be easier with current computer facilities.  

I

2009/10/30 Brian G. Peterson <[hidden email]>

John,

Thanks for the pointers.  I'll certainly research this method.
It appears on first reading that the Chow-Lin method requires multivariate indicator series to attempt to remove biases potentially introduced by the higher frequency indicators.  Do I understand this correctly?
If so, is the resulting series truly "unbiased", or are the biases introduced by construction via the indicator, and then minimized by the distribution of the residual?

Also, you would not be able to decompose a univariate series of annual numbers, as I thought was proposed by the original poster in this thread, without identifying indicators.  Correct?

Thanks again,

 - Brian

John Frain wrote:

Several Official Bodies (Central Banks Eurostat etc) use Chow-Lin
interpolation to derive  quarterly data from annual or monthly from
quarterly.  It may be the case that the higher frequency data have
only recently been produced.  I don't know if anyone has produced any
R routines but if you google "Chow-Lin interpolation" you will
probably find implementations in Matlab or Gauss that should transfer
easily to R.  My implementation in RATS can  be accessed at
ideas.repec.org/p/cbi/wpaper/2-rt-04.html.  This contains an
explanation of the methodology and references to the original papers.

The research section of the Eurostat website also contained some
relevant material.

Best Regards

John

2009/10/30 Brian G. Peterson <[hidden email]>:
 

Axel Leroix wrote:
   

Hi,
This is a newbie question. I would to be able to convert annual time
series of flow data into quarterly data. I wonder if there is any existing
R-function which permits to do it? In what package ?
 I the archive, i found that some poeple speak about "tempDis" package for
performing time series temporal disaggregation, but when I try to download
it I can not found it in the list of proposed packages.  Thank you in
advance for your help.

     

Well, as discussed multiple times on this list, going from annual (or any
lower frequency) data to quarterly (or any higher frequency) data is
questionable at best.  Think data snooping or look-ahead bias in your
modeling.

Going the other direction, from say daily (or any higher frequency)  to
monthly (or any lower frequency) , is easily accomplished with to.period for
price/value data or Return.cumulative for returns data.

If you really do want to go in the black-magic direction of going from
annual to quarterly, first make sure that the "annual" data was not first
reported as monthly data or quarterly data (this is true for almost all
macroeconomic series) and then go back to the source data at a higher
frequency.

If even this is not possible, and you insist on the highly dubious practice
of taking an annual number and turning it into four quarterly numbers, see
the various na handling methods provided by the zoo package, most likely
na.approx or na.spline.

Regards,

  - Brian

--
Brian G. Peterson
http://braverock.com/brian/
Ph: 773-459-4973
IM: bgpbraverock

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only.
-- If you want to post, subscribe first.


   

 

--
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.