Thursday, November 3, 2022
Introduction to Digital Phenotyping and the Beiwe Platform: Concepts, Design, Theory, Analysis
Richard A. and Susan F. Smith Campus Center of Harvard University, 1350 Massachusetts Avenue, Cambridge, MA
ABSTRACT
This workshop is intended for anyone with a serious interest in conducting research studies using both active data (e.g., surveys, EMA) and passive data (e.g., GPS, accelerometer, communication logs) collected from smartphones. The morning consists of lectures and the afternoon consists of hands-on lab exercises. The conceptual and theoretical content of the workshop is mostly platform independent, but the labs use data collected with Beiwe, our open-source, high-throughput platform for smartphone-based digital phenotyping. The labs also rely on Forest, our open-source data analysis library in Python, which implements a growing suite of methods for analyzing raw smartphone data. The combination of Beiwe and Forest make it possible to conduct exploratory data analyses and fit simple models with just a few lines of code. Topics that are covered in lectures and labs include inspecting raw data, assessing data volume and extent of missing data, working with surveys and survey metadata, imputing missing GPS data and studying mobility patterns, working with accelerometer data to quantify physical activity, using communication logs to assess circadian rhythm, and implementing simple longitudinal models. All workshop materials (videos, slides, data, code) will be made available to participants.
PREREQUISITES
Familiarity with the Python programming language at a basic level is a prerequisite as all hands-on exercises will be conducted using Python and Jupyter Notebooks. Additional prerequisites are general proficiency with computing and some background in statistics and probability including linear regression, logistic regression, and ideally generalized linear models (GLMs). Participants are also expected to bring their own laptop to the workshop.
INTENDED AUDIENCE
The workshop is intended for anyone with a serious interest in smartphone-based digital phenotyping. We anticipate participants to be graduate students, postdocs, and faculty as well as investigators from various research organizations and industry.
VENUE
The workshop will be held in the Richard A. and Susan F. Smith Campus Center of Harvard University located in Harvard Square at 1350 Massachusetts Avenue, Cambridge, MA 02138.
INSTRUCTORS
Emily Huang, PhD; Assistant Professor of Statistics, Wake Forest University
Ian Barnett, PhD; Assistant Professor of Biostatistics, University of Pennsylvania
Jukka-Pekka “JP” Onnela, D.Sc.; Associate Professor of Biostatistics, Harvard T.H. Chan School of Public Health
SCHEDULE (tentative)
| Time | Program |
| 8:30AM | Breakfast |
| 9:00AM | Welcome and workshop overview |
| 9:15AM | Lecture: Intro to digital phenotyping and study design |
| 10:15AM | Coffee break |
| 10:30AM | Lecture: Causal inference |
| 11:30AM | Lunch |
| 12:00PM | Lecture: Longitudinal models and missing data |
| 1:00PM | Coffee break |
| 1:15PM | Interactive lab session: Exploring data, generating features, fitting models |
| 4:45PM | Conclusion |
The workshop will be held in the Richard A. and Susan F. Smith Campus Center of Harvard University located in Harvard Square at 1350 Massachusetts Avenue, Cambridge, MA 02138.
View of the workshop venue, the Richard A. and Susan F. Smith Campus Center of Harvard University.
View of the workshop classroom on the top (10th) floor of the Richard A. and Susan F. Smith Campus Center.