Prior work has shown that environmental adversity affects cognitive development. However, it is unclear how these challenges impact brain systems, like the executive system, which underlie cognition. The current study examined the association of neighborhood socioeconomic status (SES) with executive system activation during a working memory task in the Philadelphia Neurodevelopmental Cohort.
Kristin A. Murtha
Theodore D. Satterthwaite
Bart Larsen
Bart Larsen, Adam Pines, Linden Parkes, Tyler M. Moore, Azeez Adebimpe, Aaron Alexander-Bloch, Monica E. Calkins, Allyson Mackey, David Roalf, J. Cobb Scott, Daniel Wolf, Ruben Gur, Raquel Gur, Ran Barzilay, Theodore Satterthwaite
May 2020
Data Analysis and Documentation
810336 - PNC
https://github.com/PennLINC/Murtha_Nback_Adversity
/cbica/projects/Kristin_CBF/nback_adversity
N/A
The steps below detail how to replicate all aspects of statistical analysis for this project, from sample selection to high dimensional mediation.
This script generates 4 .txt files (
grp.txt
,bblid.txt
,contrast.txt
, and design.txt) and 2 csv files containing file paths to the cope and varcope maps of each included subject. The .txt files will need to be converted into the following file formats, using FSL’sText2Vest
command:grp.grp
,contrast.con
, anddesign.mat
. These inputs are stored in a directory calledmain_model_flameo_inputs
.>
This script creates 4D time-series images of all 2b<0b contrast and varcope images, and executes the main analysis in runmode
flame1
. Output will be stored in a directory calledmain_model_flameo
. The SES result is contained inzstat5.nii.gz
.
This script transforms the 4D 2b>0b contrast image into a matrix, then splits that, along with participant’s dprime and SES scores into 2 stratified samples for testing and training analyses using CARET. This script will also perform an initial PCA of the imaging data to guide the mediation.
qsub -l h_vmem=200G,s_vmem=180G qsub_matlab.sh run_mediation.m
The submit script is found /here.
This matlab script executes an initial PCA, derives # of PMD’s from the training dataset + performs significance testing and bootstrapping. The mediators derived from the training data are then applied to the testing dataset, to test for generalizability. All results are saved in .mat files
testPDMresults.mat
andtrainPDMresults.mat
. The p-values for testing and training are saved separately in .csv files for FDR correction in R.
qsub -l h_vmem=200G,s_vmem=180G qsub_matlab.sh run_mediation.m
The submit script is found /here. Results are saved as pdm1p
and pdm2p
in both .mat and .csv files.This script uses fslmath’s
ptoz
command to convert the p-values output by matlab to zcores that are interpretable by the easythresh, and executes the cluster correction with voxel height > 3.09 and cluster probability p<0.05.