LoginSign Up
← Back to all sparks
F

funcharts

INFRA · APIS
Velocity0.0

Functional Control Charts

github.com/unina-sfere/funcharts ↗

One paper, one release: the functional control chart package ships methods as they get published.

functional-datacontrol-chartsstatistical-process-controlr-packagerobust-statisticsacademic-software
◆Current state
funcharts implements control charts for functional data — statistical process control where each observation is a curve rather than a number. Its release cadence maps one-to-one onto its authors' publication record: 1.4.0 delivered the robust multivariate framework from Capezza et al. (2022), 1.7.0 the adaptive chart from Centofanti et al. (2025) as it was accepted at Technometrics. Each release adds Phase I and Phase II functions for one named method.
◆Where it's heading
The early releases were about the package as software — vectorised tensor products, precomputed B-spline inner products, faster cross-validation — and about removing assumptions, most consequentially the assumption that functional data is represented in a B-spline basis. Since 1.4.0 the software work has stopped and the package has become a delivery vehicle for the research group's methods. Robustness came first, adaptivity second, both arriving with the paper rather than ahead of it.
◆Prediction
The next release will most likely implement whatever this group publishes next, following the same Phase I / Phase II function pairing; nothing in the feed suggests infrastructure work resuming.

◆Recent moves

  1. 1y ago

    Adaptive multivariate functional control chart added

    Adds the Adaptive Multivariate Functional Control Chart as a Phase I / Phase II function pair, published alongside the Technometrics paper that defines it. Another method in the catalogue, delivered in the package's established shape rather than changing it.

    View source ↗
  2. 2y ago

    Robust multivariate framework: filtering, imputation, and both phases

    The robust multivariate framework arrives as five functions covering the whole pipeline — robust functional PCA, a filtering step, imputation of missing functional data, and the two monitoring phases. Broader than a single chart, but still one published framework transcribed into the package's existing structure.

    View source ↗
  3. 3y ago

    All fda basis systems accepted, not just B-splines

    ⚡ SPARK

    The release that removed the package's built-in assumption about how curves are represented, and tidied its function naming at the same time. Everything since has been method delivery on top of this more general foundation.

    View source ↗
  4. 3y ago

    Vectorised inner products and cross-validation, seed argument dropped

    A performance and ergonomics release: the tensor product in fof_pc() is vectorised, inner products use precomputed B-spline values, and scalar-on-function cross-validation drops its loop. The seed argument is deprecated across the package in favour of ordinary set.seed(), and sim_funcharts() reduces simulation setup to three sample sizes.

    View source ↗