[Seminar] "Statistical Challenge and Promise: Big Data, Sampling Bias" by Prof. Yu Shen, MD Anderson Cancer Center
Speaker:ProfessorDepartment of BiostatisticsUT M. D. Anderson Cancer Center
Data from controlled randomized clinical trials (RCT) are considered to be the best source of information in cancer research. However, the strengths of the RCT can be hampered by its (possibly) limited applicability, long duration, and high cost. An alternative source of data can be found in the large observational databases and longitudinally-followed patient cohorts that have emerged. These invaluable resources present new opportunities in research to provide potential insights into cancer treatment and patient care. However, such studies are not without their own set of challenges.
The complexity of sampling mechanisms and various biases associated with prospective observational studies raise considerable statistical challenges in both the design and the data analysis. Standard analysis methods and design tools are often not applicable and, in fact, are invalid for prospective cohort studies. To address the above challenges, we need practical statistical designs and innovative analytic approaches to evaluate clinical effectiveness and healthcare interventions outside of controlled clinical trials. I will show examples of observational cohort studies and describe challenges in analyzing data from such studies. I will provide practical tools for estimating sample size, and innovative methods for analyzing time-to-event data observed from prevalent cohort studies.