[R] basic cubic spline smoothing

[spencerg]

      The best reference I know for this is something I wrote with Jim 
Ramsay and Giles Hooker:  Functional Data Analysis with R and Matlab 
(Springer, 2009).  Others may have better material. 


      After "install.packages('fda')", I suggest you try 
"system.file('scripts', package='fda')", as suggested in the Preface.  
This will point you the a subdirectory of your local installation of 
"fda" that contains files with names like "fdarm-ch01.R", 
"fdarm-ch02.R", ..., "fdarm-ch11.R".  You will likely be most interested 
Figure 9.4, sections 9.4.2 and 9.4.3, script "fdarm-ch09.R".  The script 
by itself may answer your question.  If not, you may wish to consult the 
book. 


      Hope this helps. 
      Spencer Graves


hm567 wrote:
> hm567 wrote:
>   
>> I am unsure about spar being the smoothness parameter, about where to put
>> the standard errors of the points, and about the return of the
>> smooth.spline function:
>>     
>
>   
>> Smoothing Parameter  spar= 0.5  lambda= 0.006833112 
>>     
>
>   
>> best regards,
>>     
>
>   
> Basically, the implementation based on the attached paper, for a standard
> error of points =1.0,
> the smoothing is too insensitive to the lambda smoothness parameter.
> >From 1 to almost 0.01, there is almost no smoothing... Only from 0.01 to 0
> does one start to see smoothing in action with the limit at 0 being a
> straight line.
> Note that this implementation's parameter is (1 - parameter)
>
> With R smooth.spline, 'spar' reflects well the smoothness in that:
> . at 0%, the spline interpolates
> . at 40% already, its shape is very different from the 0% one  ( for my
> implementation, they are still same )
> . at 90% it is almost a straight line
> . at 100% it is definitely a straight line
>
> This is the behavior that I wish to have.
> It seems I need to change my lambda with some transformation that is similar
> to the one in the doc of smooth.spline   (spar to lambda). Perhaps the
> reverse one. But I can't see how to do it.
>
> The other question is the standard errors. What do they correspond to in the
> doc of smooth.spline?
>
> Regards,
>   


-- 
Spencer Graves, PE, PhD
President and Chief Operating Officer
Structure Inspection and Monitoring, Inc.
751 Emerson Ct.
San José, CA 95126
ph:  408-655-4567