CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: BFF: Bayesian, fiducial, and frequentist analysis of cognitive engagement among cognitively impaired older adults Authors:  Shevaun Neupert - North Carolina State University (United States) [presenting]
Claire Growney - Washington University in St Louis (United States)
Xianghe Zhu - North Carolina State University (United States)
Julia Sorensen - North Carolina State University (United States)
Emily Smith - North Carolina State University (United States)
Jan Hannig - University of North Carolina at Chapel Hill (United States)
Abstract: Engagement in cognitively demanding activities is beneficial in preserving cognitive health. The goal was to demonstrate the utility of frequentist, Bayesian, and fiducial statistical methods for evaluating the robustness of effects in identifying factors that contribute to cognitive engagement for older adults experiencing cognitive decline. We collected a total of 504 observations across two longitudinal waves of data from 28 cognitively impaired older adults. Participants systolic blood pressure response, an index of cognitive engagement, was continuously sampled during cognitive testing. Participants reported on physical and mental health challenges and provided hair samples to assess chronic stress at each wave. Using the three statistical paradigms, we compared results from models testing levels and longitudinal changes in health and stress predicting changes in cognitive engagement. Findings were mostly consistent across the three paradigms, providing additional confidence in determining effects. We emphasize the utility of the Bayesian and fiducial paradigms for use with relatively small sample sizes because they are not based on asymptotic distributions. In particular, a fiducial paradigm is a useful tool because it provides more information than p values without the need to specify prior distributions, which may unduly influence the results based on a small sample.