This paper introduces the R package `betareg`, which provides a comprehensive set of tools for performing beta regression in the R statistical computing environment. Beta regression is particularly useful for modeling variables that take values in the standard unit interval (0, 1), such as rates, proportions, and concentration indices. The package supports both fixed and variable dispersion beta regression models, allowing for more flexible and accurate modeling of data with heteroskedasticity and skewness. The authors outline the theoretical foundations of beta regression, including the model's parameterization and link functions, and discuss the implementation details in R. They provide detailed examples and replication exercises using real-world datasets, demonstrating how to fit and interpret beta regression models, assess model diagnostics, and perform various statistical tests. Key features of the `betareg` package include: - **Model Fitting**: The main function `betareg()` fits beta regression models using maximum likelihood estimation. - **Link Functions**: Various link functions are available, including logit, probit, complementary log-log, and log-log, to handle different types of data and model specifications. - **Diagnostics**: Tools for residual analysis, such as Pearson and deviance residuals, are provided to help diagnose model fit and identify outliers. - **Inference**: The package includes methods for hypothesis testing, such as likelihood ratio tests and Wald tests, to evaluate the significance of model parameters. The paper also highlights the package's ability to handle variable dispersion, where the precision parameter (a measure of variability) is allowed to vary with the regressors, and discusses extensions to dynamic beta regression models. The authors conclude by outlining future research directions, including the handling of data with zeros and ones and the development of dynamic beta regression models.This paper introduces the R package `betareg`, which provides a comprehensive set of tools for performing beta regression in the R statistical computing environment. Beta regression is particularly useful for modeling variables that take values in the standard unit interval (0, 1), such as rates, proportions, and concentration indices. The package supports both fixed and variable dispersion beta regression models, allowing for more flexible and accurate modeling of data with heteroskedasticity and skewness. The authors outline the theoretical foundations of beta regression, including the model's parameterization and link functions, and discuss the implementation details in R. They provide detailed examples and replication exercises using real-world datasets, demonstrating how to fit and interpret beta regression models, assess model diagnostics, and perform various statistical tests. Key features of the `betareg` package include: - **Model Fitting**: The main function `betareg()` fits beta regression models using maximum likelihood estimation. - **Link Functions**: Various link functions are available, including logit, probit, complementary log-log, and log-log, to handle different types of data and model specifications. - **Diagnostics**: Tools for residual analysis, such as Pearson and deviance residuals, are provided to help diagnose model fit and identify outliers. - **Inference**: The package includes methods for hypothesis testing, such as likelihood ratio tests and Wald tests, to evaluate the significance of model parameters. The paper also highlights the package's ability to handle variable dispersion, where the precision parameter (a measure of variability) is allowed to vary with the regressors, and discusses extensions to dynamic beta regression models. The authors conclude by outlining future research directions, including the handling of data with zeros and ones and the development of dynamic beta regression models.