CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: CoCoA: A conditional correlation model with association size Authors:  Danni Tu - University of Pennsylvania (United States) [presenting]
Bridget Mahony - National Institutes of Mental Health (United States)
Maxwell Bertolero - University of Pennsylvania (United States)
Aaron Alexander-Bloch - University of Pennsylvania (United States)
Theodore Satterthwaite - University of Pennsylvania (United States)
Danielle Bassett - University of Pennsylvania (United States)
Armin Raznahan - National Institute of Mental Health Intramural Research Program (United States)
Russell Shinohara - University of Pennsylvania (United States)
Abstract: In tasks that measure cognitive function, the trade-off between speed and accuracy requires that the two be studied together. A natural question is whether speed-accuracy coupling depends on other variables, such as sustained attention. Classical regression techniques, which make different assumptions about the covariates and outcome, are insufficient to investigate the effect of a third variable on the symmetric relationship between speed and accuracy. In response, we propose CoCoA (Conditional Correlation Model with Association Size), a statistical framework that adapts parametric and semi-parametric estimation methods inspired by genome association studies to quantify the conditional correlation as a function of additional variables. We further propose novel measures of the association size, which are analogous to effect sizes on the correlation scale, while adjusting for confounding. Using neurocognitive data from the Human Connectome Project, we demonstrate that greater sustained attention in a working memory task is associated with stronger speed-accuracy coupling while controlling for age.