Statistical Inference for Same Data Meta-Analysis for Neuroimaging Multiverse Analyzes

Jeremy Lefort-Besnard, Thomas E. Nichols°, Camille Maumet°

°These authors contributed equally to this work

Table of contents

• How to cite?
• Contents overview
• Reproducing figures and tables
  • Fig. 1
  • Fig. 2
  • Fig. 3
  • Fig. 4
  • Tables

How to cite?

See CITATION.

Contents overview

This repository includes the codes necessary to reproduce the analyses described in the paper "Statistical Inference for Same Data Meta-Analysis for Neuroimaging Multiverse Analyses". The repository is structured as follows:

• CITATION/: How to cite our work.
• data/: Raw data used in the analyses.
• figures/: Summary and derived data used to create figures.
• results/: Summary and derived data used to create tables.
• src/: Source code for simulations and analyses.

Reproducing figures and tables

Please note that the "NARPS" and "HCP Young Adult" multiverse outputs must be downloaded and stored in "SDMA/data/NARPS" and "SDMA/data/HCP". "NARPS" multiverse outputs can be downloaded from NeuroVault (Gorgolewski & al, 2015)NARPS. "HCP Young Adult" multiverse output can be downloaded from Germani et al. 2023

Additionally, the following scripts must be run beforehand:

1. "SDMA/src/extract_HCP_data.py" for preprocessing and computation of the intersectional mask.
2. "SDMA/src/extract_NARPS_data.py" for preprocessing and removal of non-included participants.
3. "SDMA/src/intersectional_mask.py" to compute the intersectional mask between all NARPS team outputs.

Fig. 1

This figure is saved at "SDMA/figures/SIMULATIONS/simulations_visualisation.pdf", generated by the script "SDMA/scr/SDMA_in_generated_data.py"

Fig. 2

This figure is saved at "SDMA/figures/SIMULATIONS/J20000_K20_Corr0.8.pdf", generated by the script "SDMA/scr/SDMA_in_generated_data.py"

Fig. 3

This figure combines "SDMA/figures/NARPS/hyp1/SDMA/thresholded_map_hyp1.png" generated by the script "SDMA/scr/SDMA_in_NARPS.py" and "SDMA/figures/HCP/thresholded_map_hyp" generated by the script "SDMA/scr/SDMA_in_HCP.py".

Additionally, the result maps of the SDMA estimators in NARPS and HCP shown in Figure 3 are publicly available on NeuroVault at https://neurovault.org/collections/18197/.

Fig. 4

This figure combines "SDMA/figures/NARPS/hyp1/cluster_analysis/3clusters/Fig4.png" and "SDMA/figures/NARPS/hyp1/cluster_analysis/3clusters/per_cluster_weights_3.png" generated by the script "SDMA/scr/cluster_analysis_in_NARPS.py".

Tables

The results for all tables can be found in the files "SDMA/results/NARPS/hyp1/spatial_homogeneity/Frobenius_score_NARPS.xlsx" (with "hyp1" to be adapted depending on the hypothesis number) and "SDMA/results/HCP/spatial_homogeneity/Frobenius_score_HCP.xlsx." These files were generated by the scripts "SDMA/scr/spatial_homogeneity_in_NARPS.py" and "SDMA/scr/spatial_homogeneity_in_HCP.py," respectively.

Results from Table 1 (Difference Frobenius score) in the manuscript are located in the aforementioned files within the columns labeled 'GM+WM as Qref' and 'Norm. F.'

Results from Table 2 (Impact of regional vs. whole-brain Q for HCP results. Dice) in the manuscript are located in the aforementioned files within the column labeled 'DICE with GM&WM'

Results from Table 3 (Impact of regional vs. whole-brain Q for NARPS hypothesis 1 results. Dice) in the manuscript are located in the aforementioned files within the column labeled 'DICE with GM&WM'