Bivariate Survival and Time Series Analysis: New Regression Approaches and Applications

My research integrates the development of new statistical methodologies and theory with real-world applications, aiming to address complex statistical questions arising from complicated data sets. Much of my current and past work has centered on time-to-event (survival) data and time series data, where unique challenges demand specialized approaches. In this talk, I will present two recent projects: (1) regression modeling of bivariate survival data, and (2) quantifying the effect size of an intervention in an interrupted time series analysis. For the first project, I developed bivariate extensions of established univariate survival regression models and generalized the pseudo-observations approach to estimate regression parameters within these models. This work addresses key challenges such as censored observations and dependencies between event times. In the second project, I combined ideas from causal inference with time series regression models and quantified the effect size of an intervention in the absence of a control group. This approach tackles challenges specific to time series data, such as autocorrelation, long-term trends, and seasonal patterns. A significant application of this work involved estimating the impact of the COVID-19 pandemic on various public health outcomes.