SING method implementation and code to reproduce results in corresponding paper
Risk, B. and Gaynanova, I. "Simultaneous Non-Gaussian Component Analysis (SING) for Data Integration in Neuroimaging" arXiv link
jngcaFunctions.R - all the R functions needed for method's implementation and corresponding analyses
mCCAjointICA.R - supporting R functions implementing Joint ICA, mCCA + jICA, and SING-averaged variant via orthogonal Procrustes with scaling
CfunctionsCurvilinear.cpp - additional C++ functions for implementation of curvilinear algorithm used for SING optimization; equivalent R functions can be found in jngcaFunctions.R
makecifti.R - R function to make cifti files, modified version from Mandy Mejia, see her original version with instructions here
ComponentsForSim_setting2.Rda - Shared and individual components used to simulate large-scale data for Simulation Setting 2. Also used as input to generate sparse components for Simulation Setting 3.
community_affiliation_mmplus.csv - File to support visualization of resting state correlations, our version is modified from this paper. A similar version can be also found here.
s1_rank_simulations_small.R - Simulation setting 1, rank estimation. (WARNING: this should only be run on a cluster as it takes significant time due to a total of 400 evaluations, and 1000 permutations for rank selection)
s2_estimation_simulations_small.R - Simulation setting 1, component and mixing matrices estimation performance. Can be run in parallel with s1.
s3_results_analysis_small.R - Simulation setting 1, uses the saved results from s1 and s2 to generate figures in the paper
s4_generate_data_sim_setting2.R - Simulation setting 2, uses the components in supplied .Rda file as true components to generate simulated data.
s5_estimation_sim_setting2.R - Simulation setting 2, apply joint ICA, mCCA + jICA and SING variants, also apply rank estimation approach.
s6_results_sim_setting2.R - Simulation setting 2, use the results from s5 to calculate component and mixing matrix estimation errors, create figures in the paper.
s7_generate_data_sim_setting3.R - Simulation setting 3, uses the components in supplied .Rda file with additional truncation as true sparse components to generate simulated data.
s8_estimation_sim_setting3.R - Simulation setting 3, apply joint ICA, mCCA + jICA and SING variants, also apply rank estimation approach.
s9_results_sim_setting3.R - Simulation setting 3, use the results from s8 to calculate component and mixing matrix estimation errors, create figures in the paper.