Open dlee138 opened 9 years ago
There is a section at the end of the paper regarding limitations of their method. One of the major limitations mentioned is a decreased sensitivity to localized sets of connections that may be associated with a particular phenotype. Another thing to notice is that p-values are found for the correlation distance matrix against a created null distribution. There is dependence between the matrix for one voxel and the matrix for another, and this dependence will be reflected in the p-value, so this is something to keep in mind as well.
The Shehzad paper mentioned the benefits and the reasoning behind using MDMR in their studies, such as its ability to allow more than one predictor variable at a time. However, are there any drawbacks with this method, which we should be wary of?