THzTools: data analysis software for terahertz time-domain spectroscopy

Changed

• Change default noisefit algorithm to estimate mu using a weighted average of the measured waveforms after rescaling and shifting, rather than optimizing the cost function with respect to mu directly (#85) (Alireza Noori)
• Change procedure to evaluate parameter uncertainties, from direct computation of the Hessian to numerical evaluation using numdifftools (#85) (Alireza Noori)
• Deprecate mu_err and p_err attributes of FitResult class (#108) (Alireza Noori)

Added

• Add etfe function (Augusten Fairbairn, Japneet Sandhu, Chloe Chan)
• Add workers parameter to noisefit function to allow parallel processing of FFTs (#78) (Rabin Meetarbhan)
• Add links to top keywords in documentation (Wesley Xu)
• Add est_mu parameter to thztools (Alireza Noori)
• Add support for Python 3.14 (#92) (Steve Dodge)
• Add mu_cov, psi_cov, delta_norm, delta_norm_cov, epsilon_norm, epsilon_norm_cov, and delta_norm_epsilon_norm_cov attributes to FitResult class (#108) (Alireza Noori)

Removed

• Remove support for Python 3.9 (#92) (Steve Dodge)

Fixed

• Fix noisefit examples in documentation (#81) (Alireza Noori)
• Apply Bessel correction to estimated noise model parameter uncertainties returned by noisefit (#86) (Alireza Noori)
• Fix error in cost function Jacobian used in noisefit (#87)) (Alireza Noori)
• Fix erroneous warnings when using NumPy <2.3 on Apple Silicon M4 chips numpy/numpy#28687
• Fix error in cost function Jacobian when numpy_sign_convention = False (7c11c58) (Steve Dodge)
• Fix inconsistent internal scaling of eta parameter (c0a725f)(Steve Dodge)