Supporting code to perform the simulations and analyses of the manuscript with the same title available at https://doi.org/10.1101/631473.
To cite this repository, please cite the corresponding manuscript:
"Spatial Confidence Sets for Raw Effect Size Images" Alexander Bowring, Fabian Telschow*, Armin Schwartzman*, Thomas Nichols*. bioRxiv 631473; doi: https://doi.org/10.1101/631473
MATLAB2016b(or later) for the simulations.SPM8for the simulations.SPM12to run the permutation test for the traditional inference results displayed in the supplementary figures of the manuscript.
The Sim_??_scripts in this repository contain all the code used to obtain the 2D & 3D simulation results presented in Section 4.1, 4.2 & 4.3 of the manuscript.
In particular, the Sim_[01,..,04]_scripts directories contain the code used for the 2D simulations in Section 4.2 of the manuscript. The Sim_[05,..,11]_scripts directories contain the code used for the 3D simulations in Section 4.3 of the manuscript. The Sim_[12,13]_scripts directories contain the code used for the 2D simulations applying the Gaussian wild bootstrap method from SSS, for the methodological comparisons made in Section 4.1 of the manuscript.
In each Sim_??_scripts directory, the Sim_??.m file contains the code to generate the synthetic data, apply the confidence sets method, and then assess the empirical coverage. The Sim_??_subjects.m and Sim_??_subjects.sh files contain code to run multiple simulation jobs in parallel on a high performance computing cluster, saving each job as a .mat file. Users will need to change the SPM8 directory path in the Sim_??_subjects.m file to the appropriate directory on their own machine. Finally, the concatenate_??_subject.m files contain code to concatenate the .mat results files obtained from each simulation job into a single .mat file.
The Sim_01_scripts directory contains the code for the 2D Linear Ramp signal simulations with homogeneous variance, corresponding with the red curve results in Figure 7.1 of the manuscript.
The Sim_02_scripts directory contains the code for the 2D Linear Ramp signal simulations with heterogeneous variance, corresponding with the blue curve results in Figure 7.1 of the manuscript.
The Sim_03_scripts directory contains the code for the 2D Circle signal simulations with homogeneous variance, corresponding with the red curve results in Figure 7.2 of the manuscript, as well as the multiplier t-bootstrap results in Figure 6.1 of the manuscript.
The Sim_04_scripts directory contains the code for the 2D Circle signal simulations with heterogeneous variance, corresponding with the red curve results in Figure 7.2 of the manuscript.
The Sim_05_scripts directory contains the code for the 3D Small Sphere signal simulations with homogeneous variance, corresponding with the red curve results in Figure 8.1 of the manuscript.
The Sim_06_scripts directory contains the code for the 3D Small Sphere signal simulations with heterogeneous variance, corresponding with the blue curve results in Figure 8.1 of the manuscript.
The Sim_07_scripts directory contains the code for the 3D Large Sphere signal simulations with homogeneous variance, corresponding with the red curve results in Figure 8.2 of the manuscript, as well as the multiplier t-bootstrap results in Figure 6.2 of the manuscript.
The Sim_08_scripts directory contains the code for the 3D Large Sphere signal simulations with heterogeneous variance, corresponding with the blue curve results in Figure 8.2 of the manuscript.
The Sim_09_scripts directory contains the code for the 3D Multi Sphere signal simulations with homogeneous variance, corresponding with the red curve results in Figure 8.3 of the manuscript.
The Sim_10_scripts directory contains the code for the 3D Multi Sphere signal simulations with heterogeneous variance, corresponding with the blue curve results in Figure 8.3 of the manuscript.
The Sim_11_scripts directory contains the code for the Biobank signal simulations with Biobank variance, corresponding with the red curve results in Figure 8.4 of the manuscript.
The Sim_12_scripts directory contains the code for the 2D Circle signal simulations with the gaussian wild bootstrap, corresponding with the gaussian wild bootstrap results in Figure 6.1 of the manuscript.
The Sim_13_scripts directory contains the code for the 2D Large Sphere signal simulations with the gaussian wild bootstrap, corresponding with the gaussian wild bootstrap results in Figure 6.2 of the manuscript.
The HCP_analysis_scripts directory contains the code used to analyse the HCP data. The HCP_contour_inf.m file contains the code used to obtain the confidence sets (and yellow point estimate set) for the HCP data displayed in Figure 9 and Figure 10 of the manuscript. The permutation_test.m file is the SPM12 script used to apply the traditional inference procedure corresponding to the results in the supplementary figures for the manuscript.
The Confidence_Sets.m function contains code to obtain Confidence Sets on your own data. The function has 4 inputs: 4D_COPES, a 4D volume containing the (subject-level) effect estimate images which you would like to obtain confidence sets for. Individual effect estimate images can be concatenated into a 4D image using the fslmerge command-line tool that comes with FSL:
fslmerge -t 4D_FILENAME EFFECT_ESTIMATE_IMAGE_1 EFFECT_ESTIMATE_IMAGE_2 ... EFFECT_ESTIMATE_IMAGE_N.
GROUP_MASK_IMAGE is the group mask. THRESH is the threshold c in raw change units. OUT is the output directory where the CS images are saved.