Xiaoxuan Cai joined the The Ohio State University Department of Statistics as an Assistant Professor in 2024. She received her PhD in Statistics and has since developed a research program at the intersection of statistical methodology and mental health applications, with a particular focus on intensive longitudinal data and mobile health (mHealth) studies.

Her research is motivated by modern data streams that capture human behavior in real time, such as ecological momentary assessment (EMA) and wearable sensor data. These data present unique statistical challenges, including strong temporal dependence, non-stationarity, heterogeneity across individuals, and complex missingness patterns. Her work aims to develop principled statistical methods that can capture these dynamics and support individualized decision-making.

Her current statistical research interests include

• hierarchical Bayesian state-space models
• dynamic and individualized treatment effects
• non-stationary time series and longitudinal data analysis
• causal inference with time-varying treatments
• missing data and data integration in intensive longitudinal settings

Her current applied research interests include

• precision mental health
• suicide risk modeling and prevention
• mobile health (mHealth) and digital phenotyping
• behavioral and psychiatric data science

For students who are interested in reading courses, research assistant positions, or dissertation topics, you are welcome to contact her if you are interested in

• time series analysis 
• latent state and dynamic modeling
• individualized treatment effect estimation
• causal inference in intensive longitudinal settings
• missing data and real-world data analysis
• statistical methods for mobile and sensor data

Her research style is problem-driven. She emphasizes starting from real scientific and clinical questions, and then developing appropriate statistical tools to address them. Depending on the problem, this may involve Bayesian modeling, frequentist inference, machine learning, or hybrid approaches. The goal is to produce methods that are both theoretically sound and practically useful in high-impact application domains.