"Univariate versus multivariate methods for lesion-symptom mapping", Maria Ivanova скачать в хорошем качестве

"Univariate versus multivariate methods for lesion-symptom mapping", Maria Ivanova

Lecture in the C-STAR series, by Maria Ivanova, PhD (University of California, Berkeley), on May 21st, 2020. Website: https://aphasia.berkeley.edu Full title: Advancing our understanding of lesion symptom mapping tools: An empirical comparison of univariate versus multivariate methods Abstract: Lesion-symptom mapping tools are widely used to identify the neural structures critical for a given behavior. Some studies have claimed that novel multivariate lesion-symptom mapping methods, that consider the effects of all lesioned voxels on behavior in one model simultaneously, should be superior to traditional univariate approaches, that compare performance serially on a voxel by voxel basis. However, many of these arguments have been presented theoretically without rigorously comparing the two approaches. In my talk I will discuss our current work that aimed to address these knowledge gaps and provide a comprehensive empirical appraisal of several univariate and multivariate methods with a large stroke lesion dataset, using both synthetic and real behavioral data, across a range of relevant parameters. Our results showed no clear superiority of either univariate or multivariate methods overall, as accuracy of methods varied differentially depending on specific conditions. I will conclude by outlining limitations and advantages of different approaches and highlighting methodological implications of our findings for enhancing accuracy and robustness of future lesion-symptom mapping work.