The RBATS package is an R package that provides routines for sequential inference within the class of univariate Normal Bayesian Dynamic Models. Much of the code is based on the descriptions from West and Harrison (1997).

The main idea behind the package is that it implements Bayesian Dynamic Models, maintaining Bayesian sequential inference while keeping a lightweight structure with fewer dependencies.

In the current version, I translated the main functions, forward_filter_dlm and backward_smoother_dlm, to C++ using the Rcpp and RcppArmadillo interfaces.

You can install the development version from GitHub with:

if (!require(remotes)) install.packages('remotes')
remotes::install_github("AndrMenezes/RBATS")

I mainly wrote RBATS for use in routine data analysis and for learning purposes. I drew inspiration from the great PyBATS python package by Isaac Lavine and Andrew Cron.

• v0.2.1: Translated the main code to C++, included new non-linear DLM models and a forecast method for dlm.fit objects.

• v0.1.0: First release version written in R and used in Migon et al. (2023).

Other R packages that can be used for Dynamic Linear Model or also known State Space Models are:

• dlm by Giovanni Petris and Wally Gilks.

• bsts by Steven L. Scott.

• KFAS by Jouni Helske.

• bssm by Jouni Helske and Matti Vihola.