[R] Leave-one-out CV estimator for local polynomial regression in R?

Li Xiaoxi lixiaoxi1978 at hotmail.com
Sun Apr 23 04:17:16 CEST 2006


Hello all,

I sincerely hope someone can help me. I am wonder if there is a function in 
R that can calculate 'leave-one-out' Cross Validation estimator for 
polynomial regression? The function 'lpridge' is a pretty good one for local 
polynomial regression, but it cannot handle the leave-one-out estimation 
unless I use a loop to remove one point at a time.

Here are some details if needed. Suppose the model is y = m(x) + e, and the 
local polynomial estimation can be written as:

\hat m_i = w_i * y for i th point, where w_i is the 'smoother' vector.

The leave-one-out estimator is then \hat m(-i) = w_i' * y, where 
w_i'=w_ij/(1-w_ii) if j!=i and w_j' = 0 if j=i.

Any help will be highly appreicated.

Xiaoxi




More information about the R-help mailing list