Jiwei Zhao is an assistant professor in the Department of
Biostatistics at the University at Buffalo School of Public Health
and Health Professions.
I've always been curious about how rigorous mathematical modeling can be applied in real applications and used to facilitate an understanding of the underlying black box of the real data. Even more so, I am enthusiastic about how mathematical modeling can be used to move medical research forward and improve the health of people.
My research is data driven and that excites me. For example, longitudinal data with informative observation process can lead to follow up studies with non ignorable dropout. In addition, I am interested in using statistical methods and machine learning techniques to solve problems with missing data, high dimensional data, in causal inference and biased sampling contexts.
My research has many applicable real world applications. For example, I worked with medical doctors in the UB Department of Orthopaedics on a study to assess whether the treatment of meniscal tears with arthroscopic knee surgery was beneficial to a patient. The WOMAC pain score was used to analyze patient outcomes for a year after surgery. After following patients for one year, the doctors determined that there was a large amount of missing values due to a variety of reasons. Inappropriately addressing the missing values would result in biased results and even incorrect conclusions.
With my expertise in missing data methodology, we were able to show that the common surgical practice of treating meniscal tears does not benefit the patient because patients that did not have the dislodged cartilage removed recovered faster, with less pain and ended up a year later with identical results as if they would have had the surgery.
Since joining UB, there are several accomplishments that I am proud of.
I have worked with my PhD student to establish a novel methodology on variable selection and tuning parameter determination in statistical models under a very flexible missing data mechanism. We have applied our method to patient reported outcomes, commonly seen in biomedical studies, and the results are very promising. We are working on finalizing our procedure into an R package to better disseminate our work.