SAS and R programs:
| I find interest in every research
problem I am working on. That is why my research interest is closely
related to my research experience.
My research interests can be roughly partitioned into the following categories:
(1) using auxiliary information in statistical estimation;
(2) missing and censored data;
(3) survey data analysis (large, complex surveys).
Using auxiliary information in statistical estimation was my first research topic. I started working on this topic during my diploma work (1995-1996). When I graduated Tomsk State University (TSU), Russian Federation, I enrolled in the TSU-post-graduate program of the Department of Applied Mathematics and Cybernetics and continued working on auxiliary information problems. The original problem I focused on was "probability estimation when probabilities of some other events are known". This was relatively artificial for me. I did not see real-life problems behind the underlying mathematics. I started thinking that the additional probabilities cannot be known exactly in real-life. Instead, these probabilities could have been obtained from previous experiments. Using this practical correction, I was able to find some interesting formulas, and surprisingly for myself, I realized that I did not have strong background in statistics. My mathematical knowledge did not cover a variety of statistical methods developed for data analysis. At that time, I decided to take additional education in statistics. After I defended my dissertation for candidate of technical sciences (2002), I left for the USA and enrolled as a PhD student of the Department of Statistics at the University of Kentucky.
My interest in missing and censored data evolved out of the approaches I used in "using auxiliary information ..." that can be successfully applied to this new area. After proper development of these techniques my co-authors and I proposed an alternative method for estimating parameters in the presence of missing data. New method is distribution-free and does not require using imputations in any form. This approach can be safely applied for sufficiently large sample sizes if the missing data mechanism does not affect on the mean and variance-covariance structures. This assumption holds for larger set of problems than the assumption of ignorable missing data does. Applying the proposed approach to random dropout, the well-know Kaplan-Maier estimator was obtained. Working on change score estimation interesting results were found.
Since January 2004 I have been the statistician to the "Intimate Partner Violence Surveillance Project" (funded by Centers for Decease Control and Prevention) at the Kentucky Injure Prevention and Research Center. During five years (1999-2003) simple random samples of about two thousand Kentucky women each year were interviewed about intimate partner violence. The analysis of these surveys is my responsibility. The results of these analyses are presented in a number of technical reports, peer reviewed articles, training materials and violence prevention initiatives.
Separately, I should say, I like programming. People with Russian education in applied mathematics usually become programmers. I do not have problems with new programming languages. It is easy to learn them, if you know how they are supposed to work. Three years ago, my favorite languages were C++ (I used STL) and Visual Basic for Applications. But now, because of my strong interest in statistics, I program on SAS and R. I do take great pleasure solving programming problems. This work is both challenging and exciting. For example, working on survey analysis, I like to write IML-based SAS-macros from time to time. It is not required, but it makes routing work much more interesting.