Code for the numerical experiments in Section 4.1 of the paper 'Robust Geodesic Regression' by Ha-Young Shin and Hee-Seok Oh.
The 'GeodRegr' package is required, as are 'MASS', 'mvtnorm', 'R.matlab' and 'rmatio'.
-
simulation_experiments_1.R
contains the code for the simulation experiments on the sphere$S^n$ and hyperbolic space$\mathbb{H}^n$ for type$N$ ,$T$ and$C$ errors. The output is the MSEs for$p$ and$v^j$ . Under initializations, the dimension, the number of independent variables$k$ and the type of M-type estimator can be changed. -
simulation_experiments_2.R
contains the code for the simulation experiments on the sphere$S^n$ and hyperbolic space$\mathbb{H}^n$ for various values of$\sigma$ . The output is the sample variance for$p$ . Under initializations, the dimension, the number of independent variables$k$ and the type of M-type estimator can be changed. -
kendall_experiments.R
contains the code for the experiments on Kendall's 2-dimensional shape space. The type of M-type estimator can be changed under the initializations section. We have used the preprocessed data provided by Cornea and Zhu on their website under the title 'Regression Models on Riemannian Symmetric Spaces'.