Using Multiple Imputation, Survival Analysis, And Propensity Score Analysis In Cancer Data With Missingness

In this thesis multiple imputation, survival analysis, and propensity score analysis are combined in order to answer questions about treatment efficacy in cancer data with missingness. While each of these fields have been studied individually, there has been little work and analysis on using all three together. Starting with an incomplete dataset, the goal is to impute the missing data, and then run survival and propensity score analysis on each of the imputed datasets to answer clinically relevant questions. Along the way, many theoretical and analytical decisions are made and justified. The methodology is then applied to an observational cancer survival dataset of patients who have brain metastases from breast cancer to determine the effectiveness of chemotherapeutic and HER2-directed therapies