NEWS | R Documentation |

Function

`ppKernel`

added, which implements non-parametric estimation of filter functions in reproducing kernel Hilbert spaces.Function

`ppSmooth`

replaces`pointProcessSmooth`

(with a few changes in the argument list, but quite a few changes in the internals). It implements non-parametric estimation of smooth filter functions using B-spline basis expansions.Function

`ppLasso`

added. It relies on`glmnet`

.

Argument

`fixedCoefficients`

for`pointProcessModel`

has been removed. In future versions the formula interface for model specification will support the use of`offset`

, but for the current version there is no longer support for fixed components in the linear predictor.Argument

`Omega`

for`pointProcessModel`

has been replaced by`lambda`

. The function no longer supports general quadratic penalty matrices but only diagonal penalization.

The class

`MultivariatePointProcess`

has been introduced, which contains a list of`PointProcessModel`

s. Each element in the list is a model of one coordinate given one or more of the other coordinates, and in total the list comprises a multivariate model of point processes.Methods such as

`summary`

,`termPlot`

,`stepInformation`

and`ppmFit`

have been implemented for the`MultivariatePointProcess`

class.The

`pointProcessModel`

function now interprets a vector of variables on the left hand side of a formula as the specification of a multivariate point process model in which case the function returns an object of class`MultivariatePointProcess`

.

The implementation of the linear filters has been modified slightly concerning the treatment of the boundary.