Dataset: Methods for computing the maximum performance of computational models of fMRI responses.

Accompanying data for the revised version of the manuscript: Methods for computing the maximum performance of computational models of fMRI responses. written by Agustin Lage-Castellanos, Giancarlo Valente, Elia Formisano, Federico De Martino, submitted for publication in Plos Computational Biology, November 2018. The dataset (01.rar) contains the fMRI time series for one subject in Nifti format acquired on an actively shielded MAGNETOM 7T whole body system driven by a Siemens console at Scannexus (www.scannexus.nl). Every folder (24 runs, one folder per run) contains 150 Nifti files, one Nifti file for each fMRI volume. Preprocessing consisted of slice scan-time correction (with sinc interpolation), 3-dimensional motion correction, and temporal high pass filtering (removing drifts of 4 cycles or less per run). The matlab file dmS01_24runs.mat contains a 24-length cell array of fMRI design matrices, one for every run. Every fMRI design matrix is size 150 volumes x 51 covariates. The first 42 columns correspond to the stimuli presented (42 sounds per run). Columns 43, and 44, correspond to the run mean and the linear trend covariates. The rest of the columns correspond to the covariates obtained with GLMdenoise. The matlab variable <em>stimulus</em> of size 24 x 42 contains the index of the sounds presented at every run. A total of 168 sounds were presented, each sound was presented 6 times across the 24 fMRI runs. The file SPMgls0.rar contains the Beta images in Nifti format for every column of the fMRI design matrix, including noise covariates, for every fMRI run. This model was estimated assuming i.i.d fMRI noise (OLS). The codes for computing the noise ceiling are available in the file nccodes.rar, together with a two of examples of their use. SPM is required.