Skip to content
Prev 417 / 1340 Next
Original Raw

mgcv 1.4 on CRAN

Simon Wood 
mgcv 1.4 is now on CRAN. It includes new features to allow mgcv::gam to fit
almost any (quadratically) penalized GLM, plus some extra smoother classes.

New gam features
-------------------------

*  Linear functionals of smooths can be included in the gam linear predictor,
allowing, e.g., functional generalized linear models/signal regression,
smooths of interval data, etc.

* The parametric component of a model can be quadratically penalized, giving
easy access to gam's fitting and smoothing parameter selection methods, for
any model with a penalized glm structure.

* Smooths can be linked to have the same estimated smoothing parameter.

* `by' variables (used for varying coefficient models) can now be factor
variables, to enable easy conditioning of smooths on factors.

* The default p-values for smooth terms have been substantially improved.

* see ?gam.models and ?summary.gam for further details.

New smoothers
----------------------

* Eilers and Marx style P-splines are now built in, along with a cyclic
version. See ?p.spline.

* An adaptive smoother class has been added. See ?adaptive.smooth.

* The interface for adding user defined smooths has been modified and
simplified. See ?smooth.construct.

A fuller list of changes is at
http://cran.r-project.org/web/packages/mgcv/ChangeLog

--
-------------------------------------------------------