NEWS.md

In squat: Statistics for Quaternion Temporal Data

squat 0.4.0

New features

• Adaptation to new API of fdacluster.
• Added rotation invariant elastic distance.
• Added no alignment option in distance matrix computation via SRVF.
• Added roll-pitch-yaw representation.
• Added format()/print() pair correctly.

Minor improvements and bug fixes

• Remove dependency to purrr in favor of base R *apply() family to reduce dependency trail.
• Replace furrr with future.apply for consistency with the removal of purrr.
• Update roxygen version.
• Fix bug in predict from PCA with new data. Properly compute L2 inner product.
• Fix bug in predict from prcomp object.

squat 0.3.0

New features

• Added S3 specialization of the stats::predict() function for prcomp_qts objects.
• Added function qts2aats() which allows to switch from quaternion to axis-angle representation of rotations.
• Added usual operations +, -, * and inverse_qts() for quaternion time series using the Eigen library.

Small improvements

• Make sure quaternion geodesic mean is not flipped.
• Fix issues in PCA:
• avoid numerical overflows due to bad choice of k in gam();
• improved documentation;
• Use same number of basis functions in uni- and multivariate decompositions.
• Updated GHA scripts and README.
• Fix bug related to Rcpp following RcppCore/Rcpp#1287.

squat 0.2.1

• Add use_fence robustification option;
• Adapt to changes in fdacluster;
• Properly compute tangent spaces along mean QTS;
• Update News section of website.

squat 0.2.0

Major features:

• Added hierarichal clustering;
• Added DBSCAN clustering;
• Added distance matrix computation.

Minor improvements:

• Adapted code to match new API in fdacluster package.

squat 0.1.0

Major statistical features

• A first API proposal with a class qts and a class qts_sample for which a number of methods are properly implemented.
• Available statistical methods for QTS samples:
• random generation according to the Gaussian functional model via rnorm_qts(),
• scale(),
• mean(),
• median(),
• distance matrix computation via distDTW() (i.e. for now we use the dynamic time warping),
• tangent principal component analysis via prcomp(),
• k-means with optional alignment via kmeans().
• Added multiple ways of displaying samples of QTS.
• Added two example datasets.

Improvements

• Make all functions applicable to a single QTS also applicable to QTS samples, with appropriate class for the output.
• Enable as_qts_sample() to generate a QTS sample of size 1 from a single QTS as input argument.
• Rename change_points argument to the plot.qts() function to better reflect its flexibility.
• Added subset operator for QTS sample objects.
• Added append S3 method for QTS sample objects.
• Added hemispherize() function to remove any discontinuities in QTS due to quaternion flips.
• Any parallelization computation is now handled using the futureverse principles and, in particular, implemented through the use of the furrr package.

squat 0.0.1

• Added a NEWS.md file to track changes to the package.
• Initial version.

squat documentation built on June 8, 2025, 11:30 a.m.