Statistical issues with brain imaging localisation

Localisation of effects requires correction for multiple comparisons, for instance in brain imaging. However, popular methods such as cluster and FDR corrections have important (and typically ignored) limitations that prevent strong claims about effect location. I will report simulations to explain and quantify the issue, and propose an alternative approach for EEG and MEG data, using change point detection. This technique can be combined with a hierarchical bootstrap to produce location confidence intervals that integrate variability over trials and participants, a substantial improvement over standard group approaches that ignore entirely measurement uncertainty.

Reference: Using cluster-based permutation tests to estimate MEG/EEG onsets: How bad is it? https://onlinelibrary.wiley.com/doi/full/10.1111/ejn.16618