varycoef (version 0.2.9)

varycoef: varycoef: An R Package for Varying Coefficient Models.

Description

This package offers functions to work with varying coefficient models. Currently, it can model, estimate and predict spatially varying coefficient (SVC) models. Briefly described, one generalizes a linear regression equation such that the coefficients are no longer constant, but have the possibility to vary spatially. This is enabled by modelling the coefficients by Gaussian random fields with either an exponential or spherical covariance function. The advantages of such SVC models are that they are usually quite easy to interpret, yet they offer a very highe level of flexibility.

Estimation and Prediction

The ensemble of the function SVC_mle and the method predict estimates the defined SVC model and gives predictions of the SVC as well as the response for some pre-defined locations. This concept should be rather familiar as it is the same for the classical regression (lm) or local polynomial regression (loess), to name a couple. As the name suggests, we are using a MLE approach in order to estimate the model and following the empirical best linear unbiased predictor to give location-specifc predictions. A detailed tutorial with examples is given in a vignette; call vignette("example", package = "varycoef").

Methods

With the before mentioned SVC_mle function one gets an object of class SVC_mle. And like the method predict for predictions, there are several more methods in order to diagnose the model, see methods(class = "SVC_mle").

Examples

# NOT RUN {
vignette("example", package = "varycoef")
methods(class = "SVC_mle")

# }