orthoDr

Semi-parametric dimension reduction using orthogonality constrained optimization. Provides methods for regression, survival, and personalized dose finding, as well as a general-purpose solver for problems with orthogonality constraints.

Features

• Regression: SIR, SAVE, pHd, MAVE, and semiparametric efficient estimation (SEff)
• Survival: dimension reduction for censored outcomes
• Dose finding: personalized dose estimation via partial SAVE and direct/pseudo-direct learning
• General optimizer: ortho_optim() for any orthogonality-constrained objective
• OpenMP parallel gradient approximation

Installation

# From CRAN
install.packages("orthoDr")

# Development version
remotes::install_github("rqzhu-aide/orthoDr")

References

• Wen, Z. & Yin, W. (2013). A feasible method for optimization with orthogonality constraints. Mathematical Programming, 142(1-2), 397–434.
• Ma, Y. & Zhu, L. (2012). A semiparametric approach to dimension reduction. JASA, 107(497), 168–179.
• Ma, Y. & Zhu, L. (2013). Efficient estimation in sufficient dimension reduction. Annals of Statistics, 41(1), 250–268.
• Sun, Q., Zhu, R., Wang, T. & Zeng, D. (2017). Counting process based dimension reduction for censored outcomes. arXiv:1704.05046.
• Zhou, W. & Zhu, R. (2018+). Semiparametric efficient dimension reduction. arXiv:1802.06156.