For several years, Jeff Goldsmith has worked to advance the state-of-the-art in functional data analysis by developing methods for understanding patterns in large, complex datasets in neuroscience, physical activity monitoring, and other areas.
Working closely with clinicians and neuroscientists around the world, he and his collaborators have focused on improving the understanding skilled movements. This work involves reaching movements made by stroke patients: in these experiments, a patient's fingertip position is recorded hundreds of time per second for the duration of the reach. Dr. Goldsmith has developed new statistical methods to understand the impact of stroke on movement quality, and applied these to large, longitudinal datasets. In parallel, he has proposed methods for wearable device research, especially focusing on accelerometers. These devices can produce minute-by-minute (or even finer) resolution observations of activity for hundreds of participants over several days, weeks, or months. The methods developed include approaches for regression with activity trajectories as outcomes; for interpretable dimension reduction; and for aligning major patterns (like wake from sleep, mid-day dips in activity, and sleep onset) across subjects.
Dr. Goldsmith has worked to incorporate data science techniques for transparency and reproducibility into biostatistical analyses. Research projects are accompanied by robust, publicly available software and analytical pipelines that ensure the reproducibility of the results. This approach is informed by his work in teaching data science.
- Associate Professor of Biostatistics
Credentials & Experience
Education & Training
- 2007 Dickinson College
- PhD, 2012 Johns Hopkins School of Public Health
Goldsmith J, Huang L, Crainiceanu C M (2014). Smooth Scalar-on-Image Regression via Spatial Bayesian Variable Selection. Journal of Computational and Graphical Statistics, 23 46-64.
Goldsmith J, Scheipl F (2014). Estimator Selection and Combination in Scalar-on-Function Regression. Computational Statistics and Data Analysis, 70 362-372.
Goldsmith J, Greven S, Crainiceanu C M, (2013). Corrected Confidence Bands for Functional Data Using Principal Components. Biometrics, 69 41-51.
Goldsmith, J, Crainiceanu, CM, Caffo, BS, Reich, DS Longitudinal Penalized Functional Regression for Cognitive Outcomes on Neuronal Tract Measurements Journal of the Royal Statistical Society: Series C 61 453-469 2012
Goldsmith, J, Caffo, BS, Crainiceanu, CM, Du, Y, Reich, DS, Hendrix, CW Non-linear Tube Fitting for the Analysis of Anatomical and Functional Structures Annals of Applied Statistics 5 337-363 2011
Goldsmith, J, Bobb, J, Crainiceanu, CM, Caffo, BS, Reich, DS Penalized Functional Regression Journal of Computational and Graphical Statistics 20 830-851 2011
Goldsmith, J, Wand, MP, Crainiceanu, CM Functional Regression via Variational Bayes Electronic Journal of Statistics 5 572-602 2011