[R] Help with loess "Standard Error of the Residuals"

[peter dalgaard]

On May 1, 2012, at 07:27 , Bert Gunter wrote:

> John:
> 
> 1. This is not really an R question. You should post on statistics
> help list , e.f. stats.stackexchange.com
> 

If it is about the concrete implementation, then it is actually an R question. It probably won't harm to take a look at the code within stats:::simpleLoess, which at least brings some concepts to the surface: pseudoresiduals, trace.hat.out, etc. Some bells are heard ringing faintly from remote areas of my memory...

The role of the weights is also quite apparent in that piece of code.

> Also, inline below.
> 
> -- Bert
> 
> On Mon, Apr 30, 2012 at 8:44 AM, John Haslett <JHASLETT at tcd.ie> wrote:
>> Dear All
>> 
>> I'm having trouble working out what exactly loess means by its "Standard Error of the Residuals"   denoted s
>> and in particular when the weights argument is invoked.
>> 
>> For example, if the weights are weights are all =1, then s^2 is nearly sum sq res/ (n -1 - 'equiv num paras')
>> 
>> If the weights are all k then s is proportional to k
>> 
>> If the weights are unequal, I don't know what it does.
>> 
>> Help directs me to very old documentation on cloess, which sheds no great light.
>> 
>> I'm sure this is written down somewhere...........
> Don't be. Lowess/loess is very old, going back to the 1970's or early
> 80's I believe. You may have to go back to (Bill Cleveland's I
> believe) original papers.
> 
> But, more to the point, lo(w)ess is algorithm, not model-based; so it
> is not exactly clear how to determine df used by the fitting and
> therefore those left for residuals. I'm not sure how accepted the
> definitions baked into R's calculations are. Again, original resources
> may be the only resource.
> 
> 
>> 
>> Thanks in advance
>> 
>> John Haslett
>> 
>> 
>> 
>> 
>> 
>> Prof of Statistics, Trinity College Dublin
>> Office phone + 353 1 896 1114
>> 
>> 
>>        [[alternative HTML version deleted]]
>> 
>> ______________________________________________
>> R-help at r-project.org mailing list
>> 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.
> 
> 
> 
> -- 
> 
> Bert Gunter
> Genentech Nonclinical Biostatistics
> 
> Internal Contact Info:
> Phone: 467-7374
> Website:
> http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm
> 
> ______________________________________________
> R-help at r-project.org mailing list
> 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.

-- 
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com