|
Density estimation, smoothing
& splines
LOESS
Locally-weighted regression for irregularly spaced
multivariate
data estimating regression curves and surfaces by a local smoothing
procedure. Manual includes application to velocity structure of spiral
galaxy. By W. Cleveland and colleagues, distributed by Statlib
and Netlib.
LOCFIT
Package for multivariate nonlinear regression and adaptive
smoothing developed at Bell Labs and based on the book `Local
Regression and Likelihood' (Springer, 1999). Similar to LOESS but with
more flexible bandwidth options; includes cross-validation and other
model assessment tools. Code available in C and within the S-plus, S
and R software environments. By J. Sun of Case Western Reserve
University.
FITPACK
Fits curves and surfaces using splines under tension.
Distributed by GAMS and Netlib.
DIERCKX
Package of smoothing spline subroutines with automatic knot
selection. Distributed by GAMS and Netlib.
GRKPACK
Nonparametric estimation of generalized linear model
regression
surfaces by fitting smoothing spline ANOVA models for Poisson and other
data, with Bayesian confidence intervals. By Y. Wang of
University of Wisconsin, distributed by Statlib.
Nonparametric
regression
Fast implementations of nonparametric curve estimators
including
local linear regression, the Nadaraya-Watson estimator and kernel
density estimators. By J. Fan of University of North Caroline,
distributed by Statlib.
Kernel density
estimation
Gaussian smoothing using fast Fourier transform.
Applied
Statistics algorithm #176.
Mixture models
Maximum likelihood estimates of mixture of normal, Poisson
or
other distributions. Applied Statistics algorithm #203.
Mixture models
Maximum likelihood estimates of mixture of normal, Poisson
or
other distributions. Applied Statistics algorithm #221.
Dip statistic
Computes Hartigan's statistic to test for unimodality.
Applied
Statistics algorithm #217.
Univariate
kernel smoother
Calculates probabilities of bins for a multinomial vector
using a
quadratic kernel with smoothing determined from cross-validation.
By J. Dong & J. Simonoff of New York University, distributed by
Statlib.
|