latent-versions provides Python implementations of statistical methods for causal inference under multiple versions of treatment. It includes mixture-of-experts models for latent-version assignment, treatment assignment models, inverse probability weighting for estimating version-specific causal effects, and code for reproducing the Monte Carlo simulation studies.
Yoshikawa, K. and Kawano, S. (2026) “Identification and Estimation under Multiple Versions of Treatment: Mixture-of-Experts Approach”. arXiv:2601.00287.
MCCA Computes the multilinear common component analysis (MCCA) based on Kronecker products of mode-wise covariance matrices. This package can perform the dimensionality reduction by estimating the basis which represent the intrinsic common structure of multiple tensor datasets.
Yoshikawa, K. and Kawano, S. (2021) “Multilinear common component analysis via Kronecker product representation”. Neural Computation, 33(10), 2853–2880. (doi:10.1162/neco_a_01425).
RVSManOpt Computes the sparse reduced-rank factor regression based on manifold optimization. This package can perform estimating the rank of the coefficient matrix, selecting the number of explanatory variables which composes factors included in the regression, and selecting the number of the factors are relevant with the response variables.
Yoshikawa, K. and Kawano, S. (2023) “Sparse reduced-rank regression for simultaneous rank and variable selection via manifold optimization”. Computational Statistics, 38, 53–75 (doi:10.1007/s00180-022-01216-5).