. Stroke is the leading cause of long-term disability in the United States, with an incidence of roughly 800,000 events each year. With the annual incidence of stroke expected to surpass 1 million U.S. adults by 2025, there will be escalating stroke-related disability and economic burdens over the next decade. Thus, there is a great clinical and economic need to develop targeted neurorehabilitative strategies that are based on better charac- terization of the motor impairment after stroke using outcomes that are detailed, objective and quanti?able. Kinematic data are densely-sampled recordings of entire movements, and provide exactly the kind of data that are needed for detailed assessments motor control following stroke. While these data are commonly col- lected, typical analyses do not take advantage of their inherent richness; instead, these analyses focus on a small number of derived summaries such as including endpoint bias and variability Our proposal develops functional data approaches for kinematic data motivated by three existing datasets. Our ?rst aims introduce statistically novel method for functional response models by incorporating covariates into registration, mean regression and covariance modeling. We combine these analysis components, which are typically treated as distinct problems, to assess the relationship between motion speed and quality, and introduce novel techniques for variable selection.
Our second aims use neuroimaging to provide insights into the anatomical processes underlying motor control impairments, and consider stroke location, structural con- nections, and increased cortical activity as drivers of impairment and recovery. Throughout, we use a Bayesian approach that jointly models all parameters of interest. All new methods will be implemented in robust, publicly available software, be validated on simulated datasets designed to mimic real-data scenarios, and be deployed on the motivating datasets to generate insights into the mechanisms behind motor control impairment following stroke.

Public Health Relevance

. We propose to develop models that address gaps in the statistical literature exposed by data produced in experiments using kinematic data to assess motor control, skill, learning, and recovery following stroke. In such experiments, subjects make repeated motions which are recorded in their entirety, producing a rich dataset that allows unique insights into motor control. Analyses in the neuroscience literature has to date focused on simple summaries of this data, reducing hundreds of motions to single numbers. In place of this immense reduction we propose a collection of models using a functional data analytic perspective to provide a comprehensive framework for the analysis of such data.

National Institute of Health (NIH)
National Institute of Neurological Disorders and Stroke (NINDS)
Research Project (R01)
Project #
Application #
Study Section
Biostatistical Methods and Research Design Study Section (BMRD)
Program Officer
Chen, Daofen
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Columbia University (N.Y.)
Biostatistics & Other Math Sci
Schools of Public Health
New York
United States
Zip Code
Backenroth, Daniel; Goldsmith, Jeff; Harran, Michelle D et al. (2018) Modeling motor learning using heteroskedastic functional principal components analysis. J Am Stat Assoc 113:1003-1015
Wong, Aaron L; Goldsmith, Jeff; Forrence, Alexander D et al. (2017) Reaction times can reflect habits rather than computations. Elife 6:
Cortes, Juan C; Goldsmith, Jeff; Harran, Michelle D et al. (2017) A Short and Distinct Time Window for Recovery of Arm Motor Control Early After Stroke Revealed With a Global Measure of Trajectory Kinematics. Neurorehabil Neural Repair 31:552-560
Goldsmith, Jeff; Schwartz, Joseph E (2017) Variable selection in the functional linear concurrent model. Stat Med 36:2237-2250