[R] Kalman Filtering?

[Prof Brian Ripley]

On Wed, 15 Jun 2005, Spencer Graves wrote:

> 	  Thanks much.  I modified the R code for "KalmanLike" as you
> suggested, and the problem went away.
>
> 	  Next question:  Should KalmanLike work the way it does without your
> suggested modification?  If not, how do we get your proposal transmitted
> to the appropriate person so it is appropriately considered for adoption
> in a future release of R?

It works as it is documented to do, for a very good reason.
I've reiterated the comment in the code in the help page, for those who 
don't read the sources (like you: perhaps you should learn to).

> Thomas Lumley wrote:
>
>> On Wed, 15 Jun 2005, Spencer Graves wrote:
>>
>>
>>> 	  1.  The function "KalmanLike" seems to change its inputs AND
>>> PREVIOUSLY MADE copies of the inputs.
>>
>>
>> The strange thing is that it changes its inputs. It would be expected that
>> this also changed previously made copies. The copies will just be
>> references to the same actual object in memory (which is why changing the
>> inputs is undesirable).
>>
>> Looking at the C code, it does change the a, P, and Pnew components of the
>> model.  I'm a little reluctant to mess with this code, but an R-level
>> work-around should be to replace mod$P with mod$P+0 and so on in the
>> call to .Call, as this will force copying.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595