MICHAEL GRAY DISSERTATION DEFENSE

Title:

Creating Regression Models for non-Markov Transition Probability using Pseudo-Observations

Abstract:

Multi-State model is a graphical tool widely used to illustrate transitional relationship between states in many applications. The illness-death model is one particular example of a multi-state model.We investigated transition probabilities using a counting process approach. Aalen-Johansen estimator is the gold-standard in estimating a transitional probability. However, the Aalen-Johansen estimator is susceptible to being biased when the Markov assumption is violated. Several papers have published non-parametric estimator that accommodate for non-Markov models.

Furthermore, there are only few existing work in creating a regression model for transition probabilities in the non-Markov setting. In creating the regression model, we would like to utilize the jackknife method, pseudo-observations. An important requirement in using pseudo-observation is that we need an unbiased estimator. The Aalen-Johansen estimator could be a poor choice due to it being susceptible to bias. We propose using an alternative estimator for the pseudo-observation in creating the regression model. We will illustrate the regression model using a simulation study and in the liver cirrhosis dataset.