[R] Adaptive kernel density was ... Fw: (no subject)

roger koenker rkoenker at uiuc.edu
Sun Jul 12 22:39:46 CEST 2009


On Jul 12, 2009, at 2:56 PM, David Winsemius wrote:

>
> On Jul 12, 2009, at 3:21 PM, maram salem wrote:
>>
>> Dear group,
>> Thank u so much 4 ur help. I've tried the link,
>> http://finzi.psych.upenn.edu/R/library/quantreg/html/akj.html
>>
>> for adaptive kernel density estimation.
>> But since I'm an R beginer and the topic of adaptive estimation is  
>> new for me, i still can't figure out some of the arguments of
>> akj(x, z =, p =, h = -1, alpha = 0.5, kappa = 0.9, iker1 = 0)
>> I've a vector of 1000 values (my X), but I don't know how to get  
>> the Z
>
> That does seem rather trivial. According to the help page, those are  
> just the points at which the density should be estimated. The  
> example in the help page shows you how to create a suitable vector.
>
>> and what's Kappa?
>
> Not so obvious. Experimentation shows that reducing kappa makes the  
> estimates less smooth.
>
>> I'm sorry if the question is trivial but I hope u could recommend  
>> some refrence if u know one.
>
> Koenker gives two references and apparently you have some other  
> material you are reading. Your university should have access to the  
> Project Euclid Annals of Statistics copies that are found with the  
> obvious Google search strategy.  Maybe you should be questioning the  
> overall strategy of using a function you don't understand. Why, for  
> instance, do you even have an interest in this function?
>
The Silverman book on density estimation is as close to a canonical  
reference as one is
likely to encounter in statistics, the akj() function implements  
Silverman's adaptive kernel
proposal so it would be quite helpful to have this reference at hand.   
That is why it was
cited in the man page for the function.




More information about the R-help mailing list