Research

I am involved in a large-scale collaborative research aiming to find cure or disease modifying treatment for Alzheimer's disease, by re-purposing of existing FDA-approved drugs to Alzheimer's disease. As part of this project, I am working on development of privacy-preserving statistical methods for integrative (federated) analysis of several Electronic Health Record databases originating from the UK, US and Israeli populations.

Another research problem that I am working on concerns incompletely observed data sampled with bias and, in particular, non-probability samples. This research project is motivated by the Adult Perthes Study data, collected through a Web survey. The study was launched by the International Perthes Study Group with the aim to study what implications Perthes disease (a rare childhood disorder in a hip) has in long term, i.e.,  when the kids, who had Perthes disease, transition  into adulthood.  Although, a Web survey is an efficient way to reach out to people with a rare condition, naïve analysis of a self-selected cross-sectional sample might be misleading and can result in wrong conclusions. One of my goals is to develop methods that account for selection biases, arising from this study design. 

One of my other research goals is to develop robust methods for analysis of dependently truncated data. Left-truncated survival data arise as a result of prevalent sampling. Usually the truncation time (or the time of delayed entry into a study) is assumed to be independent of the lifetime of interest. Recently, however, the dependence between the truncation time and the target lifetime has been recognized in many newly collected big clinico-genomic databases and biobanks. However, there is only a handful of available methods that can adjust for dependence between the truncation time and the lifetime of interest.