Investigating Causality in Mobile Health Data through Deep Learning Models.

Mobile health has emerged as a practical alternative in treating and managing one’s health problems. However, most of the mobile health data are observational data collected through sensors, which makes it difficult to analyze the causality of the delivered interventions through standard regression methods. In this work, we review deep learning models that can be used to estimate the causal effect in raw mobile health data. These models are capable of handling multivariate time series data in estimating the unbiased causal effect given a sequence of treatments.