[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: RE: kernreg1
"Nick Cox" <firstname.lastname@example.org>
st: RE: kernreg1
Wed, 17 Oct 2007 22:16:44 +0100
If you are interested in nonparametric regression,
you can find much better programs than -kernreg1-.
The original -kernreg- was published in the STB in
1996. Various other authors have made public bug
fixes to -kernreg-, in -kernreg1- and -kernreg2-,
that are available from SSC. The first-named
authors in each case (Xavi Ramos and myself)
took responsibility for their changes but
also named the original authors as nominal co-authors.
Despite several public and private requests
(and promises) the original authors have never
updated their -kernreg- program. Thus we have deadlock,
as (in particular) I see some small benefit, although
it is declining asymptotically towards zero,
in -kernreg2- remaining available, as it flags
that -kernreg- is broken. It is not for me
to withdraw -kernreg1-.
That's not your question, but it's background.
The main point is that long since we have
much more versatile and up-to-date programs
for this, notably -locpoly- (now -lpoly-).
(-search- for locations.)
Also, it is very easy through various
official and user-written programs to
use any number of flexible (pun intended)
spline-type or programs for other
You could spend a year reading the
literature. The book by Ruppert, Wand
and Carroll on _Semiparametric regression_
(CUP 2003) is one of several nice accounts.
Ana R. Rios
> I have two questions about kernreg1 that I would like
> to ask: (1) how to obtain the value of the bandwidth
> in the case of optimal bandwidth? (2) is there any
> rule to follow while choosing npoint?
> Could anyone suggest a general reference on applied
> nonparametric regression? (I have just started working
> on this type of analysis).
* For searches and help try: