preprints.org > biology and life sciences > neuroscience and neurology > 202008.0428.v2

Working Paper Review Version 2 This version is not peer-reviewed

Version 1 : Received: 15 August 2020 / Approved: 20 August 2020 / Online: 20 August 2020 (05:42:04 CEST)
Version 2 : Received: 22 September 2021 / Approved: 23 September 2021 / Online: 23 September 2021 (11:13:08 CEST)

We discuss what we believe could be an improvement in future discussions of the ever-changing brain. We do so by distinguishing different types of brain variability and outlining methods suitable to analyse them. We argue that, when studying brain and behaviour data, classical methods such as regression analysis and more advanced approaches both aim to decompose the total variance into sensible variance components. In parallel, we argue that a distinction needs to be made between innate and acquired brain variability. For varying high-dimensional brain data, we present methods useful to extract their low-dimensional representations. Finally, to trace potential causes and predict plausible consequences of brain variability, we discuss how to combine statistical principles and neurobiological insights to make associative, explanatory, predictive, and causal enquires; but cautions are needed to raise association- or prediction-based neurobiological findings to causal claims.

Analysis of variance; Variance-decomposition; The Bayesian brain; High-dimensional data; Association; Explanation; Prediction; Causation; The neural law of large numbers

Biology and Life Sciences, Neuroscience and Neurology

* All users must log in before leaving a comment