Collinearity and Bias in fMRI Studies: Discussion of Mumford et al. Preprint скачать в хорошем качестве

Collinearity and Bias in fMRI Studies: Discussion of Mumford et al. Preprint

Recording from the University of Michigan's monthly NII-Methods meeting, discussing a preprint by Mumford et al.: https://www.biorxiv.org/content/biorx... Many of the slides on collinearity and Variance Inflation Factor are taken from the AFNI website and also from Jeanette Mumford's slides from the University of Michigan fMRI Course. This paper examines different ways that collinearity is addressed in fMRI studies, and how these methods can either decrease or inadvertently increase the variance inflation factor (VIF), which can bias parameter estimates. In particular, this is applied to simulations of the Monetary Incentive Delay (MID) experiment, which elicits the anticipation of a reward or loss, and also the outcome of that reward or loss. Table of Contents ============== 0:00 Overview of Collinearity and Variance Inflation Factor 16:26 Motivation for the current preprint 19:04 Different scenarios for reducing VIF 27:18 Overview of the MID experiment 32:51 The MID dataset and simulation dataset: ABCD model vs. Saturated model 37:38 Reaction time differences in conditions and potential confounds 40:33 Neuroimaging results showing potential areas of bias between ABCD and Saturated models 42:22 Summary 47:25 Discussion