INAr R Project

Generation, estimation and testing of Integer Autoregressive models

Overview

The INAr is a package for the study of integer-valued autoregressive models, namely INAR(p), considered the counterpart to the conventional autoregressive models AR(p). INAR(p) models are proved to useful for the study of realizations of random variables arising from counting, with range contained in the discrete set of non-negative integers. The package aims to provide tools for the generation, estimation and testing of these models. For a detailed description of the package functionalities, please refer to the vignette.

INAr is not the only R package for the analysis of integer-valued time series, but it is among the few ones specifically focused on INAR(p) processes. Here are some others:

• tsinteger by Manoel Santos

• spINAR by Maxime Faymonville

• tscount by Tobias Liboschik

Installation

# Install from CRAN 
# !!!---not available at the moment---!!!
# install.packages("INAr")

# Or the latest stable GitHub version
# install.packages("devtools")
devtools::install_github("blog-neas/INAr")

# Or the development version from GitHub
# install.packages("devtools")
devtools::install_github("blog-neas/INAr", ref = "devel")

Roadmap

Main Steps

The INAr package will provide functions for the generation, estimation and test INAR(p) processes with different types of innovations. Some functionalities are already available, while others are planned for the future develpoments. In particular, the main steps are listed below:

1. Generation of INAR(p) processes

• Generation of INAR(p) process with different innovations
  • Poisson
  • Negative Binomial
  • Generalized Poisson
  • Katz
• Include the possibility to use custom innovations

2. Estimation of INAR(p) processes

• CML estimation of INAR(p) processes
  • with p = 1 and Poisson innovations
  • with p = 1 and Negative Binomial innovations
  • with p = 1 and additional innovations (Generalized Poisson, Katz, Good, ...)
  • with p > 1 and additional innovations
• CLS estimation of INAR(p) processes
  • with p = 1 and Poisson innovations
  • with p = 1 and Negative Binomial innovations
  • with p = 1 and additional innovations (Generalized Poisson, Katz, Good, ...)
  • with p > 1
• SP estimation of INAR(p) processes
  • with p = 1 and Poisson innovations
  • with p = 1 and Negative Binomial innovations
  • with p = 1 and additional innovations (Generalized Poisson, Katz, Good, ...)
  • with p > 1 and additional innovations
• YW estimation of INAR(p) processes
  • with p = 1 and Poisson innovations
  • with p = 1 and Negative Binomial innovations
  • with p = 1 and additional innovations (Generalized Poisson, Katz, Good, ...)
  • with p > 1

3. Testing for the presence of INAR structure

• Sun & McCabe Test
  • Exact test for different distributions of the innovations (Poisson, Negative Binomial, Generalized Poisson and Katz)
  • Semiparametric Bootstrap test
  • Parametric Bootstrap test - Poisson, Negative Binomial and Generalized Poisson
  • Parametric Bootstrap test - Other innovations (Katz, Good, ...)
• Harris & McCabe Test
  • Exact test for different distributions of the innovations
  • Bootstrap test
• Zero Inflation Test
  • Puig and Valero (2006)
  • van den Broek (1995)
• Under- and Over- Dispersion Test (Fisher, 1950)
• Goodness-of-fit tests
  • Chi-squared test (Weiss et al, 2019)
  • Poissonity test (under development)

4. Visualization and summary of INAR(p) models

• Visualization
  • Summary
  • Fitted values
  • Plotting

Secondary Steps

• Define package structures and states
  • Functions
  • Dependencies list
• Licensing: GPL-3
• Testing and functions coverage
  • Unit tests
  • Code coverage
• Documentation
  • Function documentation
  • Vignettes
• Maintenance and distribution
  • Continuous integration
  • Releasing to CRAN
  • Lifecycle
  • References

Further steps and developments

In future developments, the package can include (but is not limited to) the following aspects:

• Forecasting INAR(p) processes

• New types of INAR models (e.g., threshold INAR, MINAR, INHAR, etc.)

• Additional estimation methods (e.g., Bayesian estimation, etc.)

• More advanced diagnostic tools and tests

Contributing to INAr development

First of all, thanks for considering contributing to INAr, contributions are welcome! INAr is an open source project maintained by people who care, and an help is always appreciated. 😊 If you are interested in improving or extending the package, feel free to get in touch. You can contact directly one of the package mantainers to discuss ideas, report issues, or propose new features.

There are several ways you can contribute to this project, you can also open an issue or submit a pull request via GitHub.

• Using INAr for a paper you are writing? Consider citing it.

• Did you discover a bug? That's annoying! Don't let others have the same experience and report it as an issue on GitHub

• Have an idea for a new INAr feature? Suggest it as an issue on GitHub.

---

Please note that this project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.