GeoBosh / GKF

An R package providing a fast and flexible implementation of the Kalman filter which handles missing values and non-positive definite variance matrices

This is a fast and flexible implementation of the Kalman filter, which can deal with missing values and non-positive definite matrices for the variance of the disturbances of the measurement equation. It is mostly written in C++ and relies fully on linear algebra subroutines contained in the Armadillo library. Due to the speed of the filter, the fitting of high-dimensional linear state space models to large datasets is feasible. The package also treats nonlinear and non-Gaussian models and allows signal smoothing and sampling from the (signal) posterior distribution.

Installing GKF

You can install the development version of GKF from Github:

library(devtools)
install_github("GeoBosh/GKF")

See the vignette.

The package passes the CRAN quality control checks, if there is interest, we would publish it there.