Researchers from Massachusetts General Hospital have found that a commonly used software package for fMRI data generates false-positive rates as high as 70%, questioning 40,000 studies.
Researchers compiled resting-state fMRI data from 500 healthy controls, dividing brain activity into voxels and sorting through them to find clusters with similar activity.
Researchers discovered a bug in the AFNI software package, reducing false positives by over 10%. However, most data from faulty studies is unavailable, making reanalysis difficult.