Permutation Test Data

Permutation Test Data

     
     We provide 100 maps of p-values generated through permutation tests (10,000 permutations) of publicly shared data from two published task fMRI neuroimaging studies.
     
The aim of the ds001 study [(Schonberg et al., 2012)](https://doi.org/10.3389/fnins.2012.00080) was to understand the neural basis of naturalistic risk-taking by having 16 healthy adult subjects participate in a Balloon Analog Risk Task. On each trial, subjects were presented with a simulated balloon and offered a monetary reward to ‘pump' the balloon. With each successive pump the money would accumulate, and at each stage of the trial, subjects had a choice of whether they wished to pump again or cash-out. After a certain number of pumps, which varied between trials, the balloon exploded. If subjects had cashed-out before this point they were rewarded with all the money they had earned during the trial, however, if the balloon exploded all money accumulated was lost.  

     The aim of the ds109 study [(Moran et al.,2012)](https://doi.org/10.1523/JNEUROSCI.5511-11.2012) was to investigate the ability of different people from different age-groups to infer the mental state of others. A total of 43 had acceptable data for the false belief task - 29 younger adults and 14 older adults. Participants listened to either a 'false belief' or a 'false photo' story. During the false belief task, participants answered questions about stories that referred either to a person's false belief (mental trials) or too outdated physical representations such as an old photograph. (physical trials). Afterward, participants had to answer a question about one of the character's perceptions of the location of the object. Then a contrast map of false belief versus false photo activations of the young adults was generated.  

     Both studies were preprocessed and analyzed in FSL using the FMRI Expert Analysis Tool (FEAT, v6.00). For each analysis, at the first level, a separate .fsf file was created for each scanning session. Runs were then combined as part of a second level fixed-effects model, yielding results which were subsequently entered into a group analysis. The preprocessing involved coregistration of the functional data to the structural brain images and then to the MNI template. In order to regress out motion-related fluctuations in the BOLD signal, six motion regressors were included in the analysis. Brain extraction was conducted on the structural image using BET.  

   Raw data downloaded from OpenfMRI.org were preprocessed using scripts available [here](https://github.com/NISOx-BDI/Software_Comparison). In the ds001 study, the first two volumes were discarded using a highpass-filter set to a sigma of 50.0s. At the run level, each of the events was convolved using a canonical double-gamma hemodynamic response function. At the subject level, the analysis of the functional data was conducted using a general linear model within FEAT where the selection of the regressors was orthogonalized. The three scanning sessions for each participant were carried out separately and then combined together. In the ds109 study, the preprocessed data was entered into a GLM for first level analysis where trials were modeled using a block and convolved using a Double-Gamma HRF from FEAT.  
    
The contrast maps for each subject were entered into a group-level one-sample t-test where clusterwise inference was conducted using a threshold of 0.01 and then a 5% FWE corrected thresholds were computed by permutation using 10,000 permutations. This procedure was done using [Randomise](https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Randomise/UserGuide#USING_randomise) from FSL.

See https://doi.org/10.1101/285585 for more details.