Pravesh Parekh, Chun Chieh Fan, Oleksandr Frei, Clare E. Palmer, Diana M. Smith, Carolina Makowski, John R. Iversen, Diliana Pecheva, Dominic Holland, Robert Loughnan, Pierre Nedelec, Wesley K. Thompson, Donald J. Hagler Jr., Ole A. Andreassen, Terry L. Jernigan, Thomas E. Nichols, and Anders M. Dale
Published in Human Brain Mapping
This repository contains the code for performing the analyses and creating figures as reported in the FEMA methods paper
-
FEMA and associated tools/utility functions Download the repository here
-
For running the applications on the ACBD Study data, you will need access to the Adolescent Brain Cognitive DevelopmentSM data
- Simulations
- Simulation 1: effect of binning and parameter recovery
- Simulation 2: comparison with
fitlme - Simulation 3: comparison of computational time
- Simulation 4: examining type I error rate
- Simulation 5: parameter recovery as a function of number of observations
- Simulation 6: type I error rate as a function of bin size
- Applications using the ABCD Study data
- Application 1: cortical thickness
- ROI-level analysis and comparison with
fitlmematrix - vertex-wise analysis
- ROI-level analysis and comparison with
- Application 2: correlation matrix dervied from resting state functional MRI
- Application 1: cortical thickness
Run sim001_doExp_effectBinning to run the experiment, followed by sim001_summarize_effectBinning to summarize the results. The script sim001_plotResults_effectBinning produces Figure 1 from the main paper while sim001_plotResults_parametersGTruth produces multiple sub-figures which were then put together as Figure 2 in the main paper.
Run sim002_doExp_comparefitlme to run the experiment, followed by sim002_summarize_comparefitlme to summarize the results. The script sim002_plotResults_comparefitlme produce supplementary Figures S5 and S6.
Run sim003_doExp_timingnObs to compare computational time between FEMA and fitlmematrix as a function of number of observations. Run sim003_doExp_timingnYvars to compare these computational time as a function of number of y variables. These are associated with their respective summarize functions sim003_summarize_timingnObs and sim003_summarize_timingnYvars. Calling script sim003_plotResults_timingnObs_FS_SE_FSE will generate Figure 3 from the main paper while sim003_plotResults_timingnObs_AE_FAE_SAE_FASE will generate supplementary Figure S7. Script sim003_plotResults_timingnYvars can be used to create Figure 4 from the main paper.
Run sim004_doExp_typeI to examine the type I error rate between FEMA and fitlmematrix. These results can be summarized by calling sim004_summarize_typeI and sim004_summarize_typeI_additionalInfo scripts, while supplementary Figure S8 can be generated by calling sim004_plotResults_typeI.
Run sim005_doExp_paramRecovery_nObs to run the experiment and sim005_summarize_paramRecovery_nObs to summarize the results. Calling script sim005_plotResults_paramRecovery_nObs will create supplementary Figures S9 and S10.
Run sim006_doExp_typeIbins to run the experiment and sim006_summarize_typeIbins to summarize the results. Calling script sim006_plotResults_typeIBins will create supplementary Figure S11.
Run abcd_makeDesignMatrix to create a design matrix that will be used for subsequent analyses.
Run abcd_CThick_analyze_ROIs to run ROI-level cortical thickness analysis. The results can be summarized using abcd_CThick_summarize_ROIs and plotted with abcd_CThick_plot_ROIs to create supplementary Figures S12, S13, and S14.
Run abcd_CThick_analyze_vertexWise to run vertex-wise cortical thickness analyses. Use the script abcd_CThick_plot_vertexWise to visualize the results (Figure 5 from the main paper and supplementary Figure S15).
Run abcd_CorrMat_analyze to run the analysis and abcd_CorrMat_plot to visualize the results (Figure 6 from the main paper and supplementary Figure S16; note that the names of the atlases/parcellation schemes were added externally).