kDGLM: An R package for Bayesian analysis of Generalized Dynamic Linear Models

Welcome to the kDGLM GitHub repository! kDGLM is an innovative R package tailored for Bayesian analysis of Generalized Dynamic Linear Models (GDLM), catering to both uni- and multivariate exponential families. This package is your go-to tool for sequential inference on time series data, offering a broad spectrum of functionalities including fitting, smoothing, monitoring, and feed-forward interventions.

Developed based on the methodology proposed in Alves et al.(2024), kDGLM integrates seamlessly with established techniques in the literature, particularly those applied in Gaussian Dynamic Models. It supports various features such as discount strategies, autoregressive components, transfer functions, and more, leveraging the Kalman filter and smoothing properties for exceptional computational efficiency. With kDGLM, you can expect almost instantaneous fitting times that scale linearly with your time series length, making it an invaluable tool for analyzing extended time series datasets.

Currently, kDGLM supports a variety of distributions, including:

• Univariate Normal (with unknown mean and observational variance);
• Bivariate Normal (with unknown means, observational variances, and correlation);
• Poisson;
• Gamma (with known shape and unknown mean);
• Multinomial (with known number of trials and unknown event probabilities).

Furthermore, kDGLM allows for the joint modeling of multiple time series, provided each series adheres to one of the supported distributions. Our ongoing development efforts are focused on continuously expanding the range of supported distributions to enhance the package's versatility.

Whether you are analyzing time series data for research, business intelligence, or any other purpose, kDGLM offers a powerful, efficient, and comprehensive solution for your needs. Explore our documentation to get started and join the community of users benefiting from the advanced capabilities of kDGLM!

Installation

Dependencies

Before installing kDGLM, ensure you have the following dependencies installed in your R environment:

• extraDistr >= 1.9.1
• cubature >= 2.1.0
• Rfast >= 2.0.8
• generics >= 0.1.3

Additionally, for an enhanced plotting experience, we recommend installing the ggplot2 and plotly packages, which augment the built-in plot methods provided by kDGLM.

Installing kDGLM

As of now, kDGLM is available directly from its GitHub repository. While we are in the process of submitting kDGLM to CRAN, you can install the latest version of the package by executing the following command in your R console:

remotes::install_github('silvaneojunior/kDGLM')

This command uses the remotes package to install kDGLM. If you do not have remotes installed, you can install it first by running install.packages("remotes").

Getting Started

After installation, load kDGLM into your R session:

library(kDGLM)

To test if the package is working properly, execute the following code exemple:

 # Poisson case
 data <- c(AirPassengers)

 level <- polynomial_block(rate = 1, order = 2, D = 0.95)
 season <- harmonic_block(rate = 1, order = 2, period = 12, D = 0.975)

 outcome <- Poisson(lambda = "rate", data = data)

 fitted.data <- fit(level, season,
   AirPassengers = outcome
 )
 summary(fitted.data)

plot(fitted.data, plot.pkg = "base")

Refer to the vignettes for detailed usage instructions. Also see the documentation of the fit method for basic examples for each type of outcome the kDGLM package offers.

Contributing

We welcome contributions! If you're interested in improving kDGLM, please consider submitting bug reports, feature requests, or pull requests.

Acknowledgements

• Raíra Marotta, who gently provided the base code for the initial versions.

Contact

For support or to provide feedback, please contact Silvaneo dos Santos Jr. or open an issue on the GitHub repository.