Skip to content


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
code and data from Latent Structure modeling paper
Shell Other
branch: master
Failed to load latest commit information.
stopwords adding stopwords


This repository contains code and relevant files for the Latent Structure
modeling project described in:

Poldrack RA, Mumford JA, Schonberg T, Kalar D, Barman B, Yarkoni T (2012). Discovering relations between mind, brain, and mental disorders using topic mapping. PLOS Computational Biology, in press.

Preprint available at:

Guide to subdirectories:
CCA - contains results from CCA analysis

NIF-Disorders - contains files used for disorders topic modeling

clustering - contains files used for clustering of disorders

cogatlas - contains files used for cogatlas topic modeling

src: contains the source files used for all analyses. these require the following
external libraries:

also note that src/utils needs to be in the python path - obtains coordinates from neurosynth database and creates images
- uses utils/ - creates merged image and computes mask of voxels with activation
on at least 1% of papers - extract full text of each article from database - read cognitive atlas concepts from RDF and get loading for each document - load NIF dysfunction ontology, grab synonyms, and add
additional missing terms - get loading for each disorder term from corpus - make documents based on cogatlas loadings - make documents based on disorder loadings - created a pickle with the full dataset to make loading easier - make files for 8-fold CV - make files for 8-fold CV - make mallet data from 8-fold files - make mallet data from 8-fold files - create scripts to run mallet jobs - create scripts to run mallet jobs - check likelihoods to get get dimenstionality - generate additional topic models to get disorder dimensionality - get dimensionality that has unique topic dists - generate scripts to run final topic models - load topic data and save loadingdata.txt - load topic data and save loadingdata.txt - make all chisquare maps - uses utils/ - make 6mm versions of topic - make slice images using p values - make slice images using p values - create latex report - create latex report - run CCA analysis

25_cluster_disorders.R - run clustering on disorders data - generate figure 2 from initial submission - generate figure 3 from initial submission - create scripts to run fulltext topic models - make 8-fold data for full text analysis - run mallet jobs on 8-fold full text - get best dimensionality for full text - get histograms of # of docs for each filtered paper - count # of locations reported across all papers - make histograms of docs/topic and topics/doc (for Figure 3) - plot empirical likelihood as function of # of topics (for Figure 2) - run CCA directly on topic distributions - get loadingdata for fold1 for each ntopics - make images using correlation rather than p value - make images using correlation rather than p value (for figure - make Figure 4 - make Figure 6 - create graphviz .dot file for topic hierarchy (Figure 5)
Something went wrong with that request. Please try again.