Multivariate Analysis is Sufficient for Lesion-Behaviour Mapping

Lesion-behaviour mapping aims at predicting individual behavioural deficits, given a certain pattern of brain lesions. It also brings fundamental insights on organization, as lesions can be understood interventions normal function. We focus here the case stroke. The most standard approach to lesion-behaviour is mass-univariate analysis, but it inaccurate due correlations between different regions induced by vascularisation. Recently, has been claimed that multivariate methods are subject lesion-anatomical bias, and move towards causal necessary eliminate bias. In this paper, we reframe problem using classical inference tools. show that, in absence additional clinical data if only one region an effect scores, suitable sufficient address This commonly encountered situation when working with public datasets, which very often lack general health data. support our claim set simulated experiments publicly available lesion imaging dataset, adequate models provide state-of-the art results.