00fRegression-package: Regression Modelling Package In fRegression: Rmetrics - Regression Based Decision and Prediction

Description

The Rmetrics "fRegression" package is a collection of functions for linear and non-linear regression modelling.

Details

 Package: fRegression Type: Package Version: R 3.0.1 Date: 2014 License: GPL Version 2 or later Copyright: (c) 1999-2014 Rmetrics Association Repository: R-FORGE URL: https://www.rmetrics.org

1 Introduction

Regression modelling, especially linear modelling, LM, is a widely used application in financial engineering. In finance it mostly appears in form that a variable is modelled as a linear or more complex relationship as a function of other variables. For example the decision of buying or selling in a trading model may be triggered by the outcome of a regression model, e.g. neural networks are a well known tool in this field.

2 Fitting Regression Models

Rmetrics has build a unique interface to several regression models available in the base and contributed packages of R. The following regression models are interfaced and available through a common function regFit. The argument use allows to select the desired model:

 1 2 3 4 5 6 7 8 9 regFit fits regression models - lm fits a linear model [stats] - rlm fits a LM by robust regression [MASS] - glm fits a generliazed linear model [stats] - gam fits a generlized additive model [mgcv] - ppr fits a projection pursuit regression model [stats] - nnet fits a single hidden-layer neural network model [nnet] - polymars fits an adaptive polynomial spline regression [polspline]

An advantage of the regFit function is, that all the underlying functions of its family can be called with the same list of arguments, and the value returned is always an unique object, an object of class "fREG" with the following slots: @call, @formula, @method, @data, @fit, @residuals, @fitted, @title, and @description.

Furthermore, independent of the selected regression model applied we can use the same S4 methods for all types of regressions. This includes, print,plot, summary, predict, fitted, residuals, coef, vcov, and formula methods.

It is possible to add further regression models to this framework either his own implementations or implementations available through other contributed R packages. Suggestions include biglm, earth amongst others.

2 Simulation of Regression Models

contains a function to simulate artificial regression models, mostly used for testing.

 1 2 regSim simulates artificial regression model data sets

3 Extractor Functions

These generic functions are:

 1 2 3 4 5 6 fitted extracts fitted values from a fitted 'fREG' object residuals extracts residuals from a fitted 'fREG' object coef extracts coefficients from a fitted 'fREG' object formula extracts formula expression from a fitted 'fREG' object vcov extracts variance-covariance matrix of fitted parameters

4 Forecasting

The function predict returns predicted values based on the fitted model object.

 1 2 predict forecasts from an object of class 'fREG'

4 Reporting Functions

For printing and plotting use the functions:

 1 2 3 4 print prints the results from a regression fit plot plots the results from a gression fit summary returns a summary report