Title: Missing data imputation via state space model for non-stationary multi-variate time series in mHealth
Authors: Linda Valeri - Columbia University (United States) [presenting]
Xiaoxuan Cai - Columbia University (United States)
Abstract: Missing data is a ubiquitous problem in biomedical and social science research. Data imputation is a commonly recommended remedy. Mobile technology (e.g., mobile phones and wearable devices) allows to closely monitoring individuals behavior and symptoms in real-time, and holds great potential for scientific discoveries and personalized treatment. Continuous data collection using mobile technology gives rise to a new type of data, entangled multivariate time series of outcome, exposure and covariates, and poses new challenges in missing data imputation for valid inference on treatment effects. Most existing imputation methods are either designed for longitudinal data with a limited number of follow-ups or for stationary time series, which may not be suitable in the field of psychiatry when mental health symptoms display dramatic changes over time or patients experience shifts in treatment regime over their course of recovery. We propose a novel imputation method based on the state-space model (SSMimpute) to tackle missingness in outcomes when multivariate time series are potentially non-stationary. We evaluate its theoretical properties and performance in extensive simulations, showing its advantages over other commonly used strategies for missing data. We apply the SSMimpute method in the analysis of a multi-year observational smartphone study of bipolar patients to evaluate the association between social network size and psychiatric symptoms adjusting for confounding.