useR! 2024

The uncertainty of model outputs is often absent or hidden in R, and tools for interacting with distributions are limited. For example, most prediction methods in R only produce point predictions by default. Although it is possible to obtain other parameters and form the complete distribution, additional knowledge about the distribution's shape and properties are needed. The distributional package vastly simplifies creating and interacting with distributions in R. The package provides vectorised distributions and supports the calculation of various statistics without needing to use shape-specific d*/p*/q*/r* functions. Statistics can be easily calculated for distributions in the same vector, regardless of shape. Manipulating distributions is also supported, including applying transformations, inflating values, truncating, and creating mixtures of distributions. When vectors are stored as data frame columns, these operations integrate seamlessly with tidyverse workflows. Distributions can also be visualised with ggplot2 using the ggdist extension package, which offers many graphical representations of uncertainty.