moveEZ: An R Package for Animated Biplots
by Raeesa Ganey, Johané Nienkemper-Swanepoel · 2026
Read moveEZ: An R Package for Animated Biplots by by Raeesa Ganey, Johané Nienkemper-Swanepoel (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-22284.txt.
Introduction
Ganey and Nienkemper-Swanepoel present moveEZ, an R package for animated PCA biplots. It shows how multivariate structure shifts across ordered levels of a category. Built on the biplotEZ package, it offers three animation frames of rising complexity. In the fixed frame, variable vectors stay put while sample points move. In the two dynamic frames, both are recomputed at each level, with Procrustes alignment keeping the view steady. The package suits publication graphics and works wherever repeated measures span an ordered category.
Catalog record
| CREATOR | by Raeesa Ganey, Johané Nienkemper-Swanepoel |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28543 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.22284
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
moveEZ: An R Package for Animated Biplots by Raeesa Ganey, Johané Nienkemper-Swanepoel Abstract The moveEZ (pronounced move easy) R package provides tools for constructing animated Principal Component analysis (PCA) biplots that represent samples and variables simultaneously. These visualizations reveal how multivariate structures evolve across the ordered levels of a categorical variable. Built as an extension to the biplotEZ package, moveEZ offers three animation frameworks of increasing methodological complexity: a fixed variable frame, in which variable vectors remain constant and only sample positions are animated; and two dynamic frames, in which both sample positions and variable vectors are recomputed and animated at each level. The dynamic frames support Generalized Procrustes analysis (GPA) for alignment and reflection to ensure visual continuity across levels, and are compatible with high-dimensional datasets including grouped structures. The package integrates with gganimate to produce high-quality animations suitable for publications and presentations, and supports both animated and static faceted displays via a single argument. Although originally motivated by tracking shifts in African climate indicators, moveEZ is domain-agnostic and applicable wherever multivariate measurements are recorded repeatedly across an ordered categorical variable, including economic, ecological, and biological settings. 1 Introduction The moveEZ R package introduces animated Principal Component analysis (PCA) biplots for the visualization of multivariate data observed repeatedly across the ordered levels of a categorical variable. PCA biplots, constructed via the companion package biplotEZ ( 10 ), are effective for revealing structure in high-dimensional data. However, when data are sequentially structured in this way, restricting visualizations to individual levels obscures the continuity of change, making gradual movement in samples or variables across the sequence difficult to detect. moveEZ addresses this by animating transitions between levels and supporting dynamic exploration, allowing analysts to perceive movement in multivariate space as a continuous process rather than a sequence of discrete comparisons. Although originally motivated by the tracking of climate conditions over time, the package is domain-agnostic and applicable wherever multivariate measurements are recorded repeatedly across an ordered categorical variable; whether that variable represents time, experimental stages, algorithmic iterations, or any other meaningful sequence. 2 PCA Biplots The aim is to approximate a high-dimensional multivariate dataset in two dimensions for visualization. Consider a dataset {\bf{X}} comprising n observations and p continuous variables, along with an additional categorical variable referred to throughout this paper as the time variable whose ordered levels define the sequential structure in the data. This variable does not need to correspond to chronological time; it may represent any ordered index such as experimental stages, algorithmic iterations, or measurement occasions. Each distinct level of the time variable partitions {\bf{X}} into a time slice: the subset of rows corresponding to that level. The collection of time slices across all levels constitutes the full sequential dataset that moveEZ animates. Where appropriate, standardization (centering and scaling) is applied prior to dimension reduction. Singular value decomposition (SVD) of {\bf{X}} yields: {\bf{X=UDV^{\prime}}}. A standard PCA biplot represents samples in principal coordinates, taken from the first two columns of {\bf{UD}} , and variables in standard coordinates, taken from the first two columns of {\bf{V}} ( 7 ). Variables in a PCA biplot can be represented either as directed vectors, as introduced by ( 3 ), or as calibrated axes, which support interpolation and prediction of sample values ( 1 , 5 , 6 ). Since moveEZ produces dynamic visualizations across multiple time points, directed vectors are preferred over calibrated axes. They produce a less cluttered display, which eases visual tracking of subtle changes in both samples and variables as the animation progresses. 3 The moveEZ framework moveEZ animates PCA biplots across the levels of an ordered variable, the time variable, by interpolating between successive biplot states using a linear easing function, ease_aes() from the tweenr package ( 14 ), applied via the transition_states() function in the package gganimate ( 13 ). Linear easing produces transitions at a constant rate, ensuring smooth and visually interpretable movement between states. A brief pause at each state allows the user to clearly identify the biplot corresponding to each time level before the animation continues. The package provides three functions: moveplot() , moveplot2() , and moveplot3() , each implementing a different conceptual approach to animating the biplot across time. The key distinction between them concerns whether the PCA solution is computed once on the full dataset or separately for each time slice, and consequently whether the variable vectors remain fixed or evolve over time. All three functions share a move argument that toggles between a dynamic animation ( move = TRUE ) and a static faceted display ( move = FALSE ), the latter being useful for figures in publications or detailed comparison across time points. Table 1 summarizes the key differences between the three functions to assist users in selecting the most appropriate approach for their data and research question. Table 1: Comparison of moveplot(), moveplot2(), and moveplot3(). Feature moveplot() moveplot2() moveplot3() PCA computed Once, on full dataset Per time slice Per time slice Variable vectors Fixed Dynamic Dynamic Sample coordinates Dynamic Dynamic Dynamic Alignment Not required Manual (align.time, reflect) Automated (GPA) Min. observations per time slice One Multiple Multiple Balance across time slices Not required Not required Required Best suited for Stable variance structure; single observations per group per time Evolving variance structure; manual alignment feasible Evolving variance structure; many time slices Table 1 shows that there is a trade-off between simplicity and advanced methodological choices which is guided by the data. moveplot() assumes that the principal components computed on the full dataset provide a meaningful basis for projecting each time slice. This is appropriate when the underlying variance–covariance structure is stable across time. When this assumption does not hold, that is, when the directional relationships among variables themselves evolve then moveplot2() or moveplot3() are more appropriate, as they compute a separate PCA solution for each time slice. The distinction between these two is one of practicality: moveplot2() requires the user to manually identify and correct orientation discontinuities between frames, while moveplot3() automates this process using Generalized Procrustes analysis (GPA), making it the preferred choice when the number of time levels is large or when discontinuities are difficult to detect visually. In all three functions, the core biplot construction is handled by biplotEZ ( 10 ), and the resulting object is passed directly to the relevant moveplot function via the pipe operator. 4 Fixed variable frame In the fixed variable frame, a single PCA solution is performed on the full dataset {\bf{X}} , yielding one set of sample scores, {\bf{Z}} , represented by plotted coordinates, and one set of variable loadings, {\bf{V}} , as vectors. The variable vectors remain constant throughout the animation, providing a stable reference frame for interpretation. The sample scores are then sliced according to the levels of the time variable and animated sequentially, highlighting how observations move relative to a fixed coordinate system defined by the overall variance structure of the data. This framework is illustrated in Figure 1 . Figure 1: The fixed variable framework where the sample coordinates in matrix {\bf{Z}} are sliced according to the levels of the time variable T and the variable vectors in matrix {\bf{V}} remaining fixed. This approach is most appropriate when the underlying variance–covariance structure can be assumed