R help - Nov 2005 - multivariate nonparametric regression with e >= 0

Hello all,

I'm a relatively new user of R, having mostly used it only for plotting so
far.  I'm also not very familiar with regression methods, hence forgive my
greenness on the topic.

What I want to do in R is multivariate nonparametric regression, with a slight
hitch.  From my experimental data I have a multitude of samples whose values
approximate a function `f' that is defined over a 5D space (i.e., f:
R^5->R).
The values of the collected samples, call these `y', approximate `f',
but due
to the process by which they are collected, they always over-estimate (i.e., y
= f + e, e >= 0).  The distribution of the error `e' can likely be
modelled
using the positive half of the normal distribution.

Naturally I'm trying to obtain a smooth and relatively faithful
approximation
of `f' using the collected samples `y'.  What would be the most fruitful
approach in R to doing this?  Even suggestions on which package/function to
use would be tremendously helpful, as I don't yet know what their
strengths/weaknesses are.

Also, I would consider parametric regression as well, but in the general case
I don't think I can assume/guess for my data at what the appropriate
parametric basis functions should be...

-- 
Maciej Kalisiak <mac at dgp.toronto.edu>
http://www.dgp.toronto.edu/~mac/