Department seminar by Yi Xiong

Please Note: This seminar will be given online.

Advancing Statistical Analyses of Event History Data

Event history data bear frequently incompleteness and heterogeneity and is thus not in the ideal format of iid observations. In this talk, I will begin with a brief review of my research on developing statistical methods for event history data with various complexly structures e.g. missing-not-at-random covariates, spatially-correlated event times with missing origins. I will then focus on tackling issues arising from administrative event history data. Administrative health data in general contain rich information for investigating public health issues. On the other hand, however, many restrictions and regulations apply to their use as administrative health databases are maintained to serve non-research purposes and only information for people who seek health services is accessible. In addition, administrative health databases evolve over time and the regulations about their access may change. Motivated by administrative records of emergency department (ED) visits by children and youths in Alberta, we propose novel statistical methods to address two challenges: (i) to evaluate dynamic pattern and impacts with doubly-censored recurrent event data and (ii) to re-calibrate estimators developed based on truncated information by leveraging summary statistics from the population. I will justify these methods theoretically and demonstrate their statistical performance using both simulation and the ED visits data. The talk will be concluded by discussion on a few future investigations.