                                 LOCFIT
                    Local Regression and Likelihood

                            Catherine Loader
                    catherine@research.bell-labs.com
                         or cathl@optonline.net
              http://cm.bell-labs.com/stat/project/locfit/

                 Also incorporating SCB code, joint with
                               Jiayang Sun
                        jiayang@sun.stat.cwru.edu
                  http://sun.cwru.edu/~jiayang/sci3.html

                             August 1, 2001


Locfit is a program for fitting regression, likelihood and density
estimation models using a local polynomial approach. Both univariate
and multivarite predictor spaces are supported, along with a wide
range of likelihoods: local least squares; local logistic regression;
hazard rate models; censored regression models; local density estimation
and others. The program can be installed as a library for R, S or S-Plus.
It can also be installed as a stand-alone application ("C-Locfit").


INSTALLATION:

To compile and install from a source distribution, please consult the
Locfit web page, http://cm.bell-labs.com/stat/project/locfit/.

If you have a binary distribution, Locfit is ready to use as distributed.
For convenience, S-Plus users may like to
  (1) move the archive to $SHOME/library/
  (2) add the line
        locfit          Local Regression, Likelihood and Density Estimation.
      to the $SHOME/library/README file.
  (3) In an S-Plus session, attach the library with library("locfit",first=T).


DOCUMENTATION:

Numerous examples can be found in my book,
"Local Regression and Likelihood", Springer Series in Statistics and
Computing, 1999.

Due to the large number of incompatable, inconvenient and changing formats
for help pages in different versions of S and S-Plus, I cannot maintain
S-Plus formatted help pages. Instead, html manual pages for most functions
are available on the locfit web pages.


ABOUT ALGORITHMS:
The Locfit implementation has been designed largely from the point
of view of generality: to provide a set of functions that can be
used for as wide a range of local fitting problems as possible.
It is not intended to be the fastest or most efficient implementation
possible. In addition to generality, the code in many places makes
extensive trade-offs made between speed and numerical accuracy (some
of which can be controlled through optional arguments). Many of Locfit's
options will only be used in a small fraction of cases; for other cases,
they add to the cost of overhead (an obvious example is multi-dimensional
fitting: when used in 1-d, many loops reduce to for(i=0;i<1;i++)).
Additionally, the user interfaces (i.e. the R and S-Plus code) add
significantly to the computational overhead. For these reasons, the
Locfit code, as distributed, should not, and can not, be used to
derive meaningful benchmarks, either for the speed or accuracy of algorithms.
Anyone wishing to use Locfit in any kind of comparative benchmark
study must contact and obtain permission from the Author.


COPYRIGHT:

Copyright (c) 1996-2001 Lucent Technologies, Bell Laboratories.
SCB code is
Copyright (c) 1996-2001 Jiayang Sun.

Permission to use, copy, modify, and distribute this software for any
purpose (with the exceptions noted in `About Algorithms' above)
without fee is hereby granted, and provided that this entire notice is
included in all copies of any software which is or includes a copy or
modification of this software and in all copies of the supporting
documentation for such software.

THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR IMPLIED
WARRANTY.  IN PARTICULAR, NEITHER THE AUTHOR NOR LUCENT TECHNOLOGIES
MAKE ANY REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE
MERCHANTABILITY OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR PURPOSE.
