Interactive calibrated-axes biplots in R with the bipl5 package

R. Buys${}^a$, S. Lubbe${}^a$ and M. L. Steyn${}^a$

${}^a$Stellenbosch University

The bipl5 package provides an interactive framework in R for constructing calibrated-axes biplots as reactive HTML widgets using Plotly and custom JavaScript. Its scope includes principal component analysis, canonical variate analysis, principal coordinates analysis, and regression biplots, combining dynamic features and embedded fit measures in a single exploratory display. Designed both as standalone software for constructing biplots and as a plotting wrapper for biplotEZ objects, bipl5 offers a flexible environment for interactive multivariate visualisation.

To address the visual clutter that occurs when calibrated axes and observations compete for the same central plotting region, bipl5 implements an algorithm that translates the axes away from the plot centre while preserving their calibration. The degree of translation can be adjusted interactively on the plot. To help users assess and manage approximation quality, the package provides predictivity-based fit measures grounded in the orthogonality framework of [1], as well as an axis-scoring procedure that evaluates each axis according to its prediction error, guiding the selection of the most informative axes for display. These diagnostics are available directly within the plot, rendering a dashboard to navigate the multivariate display.

The presentation will illustrate the functionality of bipl5 and show how classical biplot methodology can be extended into a reactive environment that combines readability, interpretability, and formal fit diagnostics.

Keywords: Biplots, Interactive visualisation, Multivariate analysis.