fdatools

Functional data analysis for Python. fdatools is a clean-room Python rewrite of R's fda package (Ramsay, Hooker & Graves): basis expansions, penalised smoothing, functional PCA and CCA, functional regression, curve registration and principal differential analysis.

• Same numbers as R. Every public function is tested against golden output from R fda 6.3.0 at rtol = 1e-8 (1e-5 for iterative fits). See the parity report.
• About 40 symbols instead of 515 functions. Curves are callable (fd(t)), arithmetic is plain Python (fd1 + fd2, fd1 @ fd2), and there is one smooth(), one register(), one .plot().
• Fits the Python stack. scikit-learn estimators, NumPy or PyTorch arrays through the Array API (gradients flow), pandas and xarray I/O.

Install

pip install fdatools

Optional extras: fdatools[plot] (matplotlib), fdatools[pandas] (pandas + xarray), fdatools[io] (read R .rds files), fdatools[torch] (PyTorch). Python 3.10 to 3.13.

Quickstart

import numpy as np
import fdatools as fdt

growth = fdt.datasets.load_growth()                   # Berkeley growth study, ships with fdatools
basis = fdt.BSpline(domain=(1.0, 18.0), n_basis=12)   # cubic B-splines on ages 1-18
coefs, *_ = np.linalg.lstsq(basis(growth.age), growth.hgtf, rcond=None)
girls = fdt.FData(coefs, basis)                       # 54 height curves in one object
speed = girls.derivative()                           # growth speed in cm/year, exact
print(speed.mean()(np.array([5.0, 12.0])))           # curves are callable: evaluate anywhere
print(fdt.inprod(girls[0], girls[0]))                 # L2 inner product of one curve

Next steps: penalised smoothing with automatic λ (smooth), functional PCA (FPCA) and more in the quickstart guide. Coming from R? The migration table maps every fda function to its fdatools equivalent.

Documentation

Full documentation, API reference and tutorials: https://hameddavodi.github.io/fdatools.

Citing

If you use fdatools in published work, please cite it together with Ramsay & Silverman (2005); see CITATION.cff.

License

BSD 3-Clause. fdatools is a clean-room implementation written from the published mathematics (Ramsay & Silverman 2005; Ramsay, Hooker & Graves 2009) and the public behaviour of R fda; it contains no R source code.