Summary: Raw code for banded ridge regression analysis of Partly Cloudy fMRI data
Hi! This is the code that was used in our paper, "Early neural development of social perception: evidence from voxel-wise encoding in young children and adults" (Im, Shirahatti, Isik. preprint)
To run this code, you will need to get the video stimulus from Richardson et al (2018) and fMRI data from Hillary Richardson's open fMRI page, download tikreg and PyMoten packages from Gallant Lab Github (https://github.com/gallantlab). To run ROI analyses, you would also need to download the ROI masks we used: We used STS mask from Ben Deen's group (https://bendeen.com/data/) and MT from Sabine Kastner's group (https://napl.scholar.princeton.edu/resources)
Broadly, there's three types of files:
- Folder 'data_prep' gives you three files: one needed to prep for the isc masking, and two for feature analysis (feature labeling csv and correlation and PCA calculation notebook)
- Folder 'rockfish_scripts' has the codes for bash and python scripts used to run the encoding model, difference analysis between age groups, and thrsholding
- Folder 'jupyter_notebooks' has the notebooks we used to visualize the data we analyzed using rockfish_scripts
- Folder 'demo_sample' gives you example of how to use the notebooks
- We also have helper files (ends with .py) where you can find the functions we used to run the codes
More specifically, to run the code, you would first want to run the rockfish_scripts folder. The workflow order should be Encoding Analysis & Differences -> Generate Null -> Threshold. After running the encoding analysis and thresholding data, you can run jupyter_notebooks folder. While all of the notebooks have interesting analysis information, the main notebooks we used for the results were figures.ipynb for whole brain analysis and ROI_indv_social_features (individual social feature model prediction), ROI_moten_all_ages (motion energy prediction), and ROI_social_all_ages (social feature combined model prediction).