Introduction

We provide programs for nonparametric and semiparametric regression function estimation. Kernel regression, local linear regression, single index regression, and additive models are included. Also, a kernel estimator and a local linear estimator of a partial derivative of the regression function is provided. In addition, linear and kernel quantile regression is implemented.

The package "regpro" can be used together with the package "denpro", which provides functions for the visualization of multivariate functions.

Package is provided by Jussi Klemelä

Installation instructions

Download the file regpro.R and then issue the command

> source("/path/regpro.R")

in the R console. For example, in Unix the command

> source("/home/aaa/regpro.R")

could work, where "aaa" is the user account. In OS X the command

> source("/Users/aaa/regpro.R")

could be used.

Contents

Kernel regression function estimation:

• pcf.kernesti: calculates a kernel regression estimator (Nadaraya-Watson estimator)
• pcf.kernesti.der: calculates a partial derivative of a kernel regression estimator (Nadaraya-Watson estimator)
• pcf.kernesti.slice: calculates a 1D slice of a kernel regression estimator
• pcf.kernesti.marg: calculates a 1D partial dependence function using a kernel regression estimator
• pcf.kern.quan: calculates estimates of the conditional quantile using kernel regression estimator on a grid of points
• kernesti.regr: calculates a pointwise estimate of a regression function using kernel regression estimator
• kernesti.der: calculates a pointwise estimate of a partial derivative of a regression function using kernel regression estimator
• kernesti.quantile: calculates a pointwise estimate of the conditional quantile using kernel regression estimator
• ma: calculates a pointwise moving average from a time series

Other regression function estimators:

• pcf.loclin: calculates an estimate on a grid of points of the regression function or of a partial derivative of the regression function
• loclin: calculates an estimate at one point of the regression function and of the gradient of the regression function
• pcf.additive: calculates an estimate on a grid of points of a regression function using backfitting in the additive model
• additive: calculates an estimate at a point of a regression function using backfitting in the additive model
• additive.stage: calculates an additive estimate at a point of a regression function using a stagewise algorithm
• pcf.single.index: calculates estimates of the regression function on a grid of points by fitting the single index model
• single.index: calculates an estimate of the index in the single index model
• linear: calculates the estimates of the coefficients of linear regression
• linear.quan: calculates the estimates of the coefficients of a linear quantile function estimate
• pp.regression: calculates a estimate at one point of the regression function using projection pursuit regression

Miscallenous:

• emp.distribu: calculates the value of an empirical distribution function
• emp.quantile: calculates the value of an empirical quantile function
• quantil.emp: calculates the empirical quantile for a level p
• copula.trans: makes a copula transform to a data matrix

Tutorial

A tutorial for the use of the package "regpro" can be found here.

Home

January 2016