My main interest is the study of change in people. More specifically, I’m interested in the ways we can model change in people statistically.
In my work as a statistician at Parnassia Group, I often get involved in projects that study the mental improvement of patients receiving mental health care. For example, if one would like to test the effectiveness of some new treatment for depression over the usual treatment, it is common practice to randomly assign a number of patients to either one of the two treatments. Before and after their treatment period, all the patients are assessed on symptoms of psychiatric distress (in our example, of course, with a special focus on depression). After collection of these data, the wonderful world of statistics enters the scene. Using statistical methods, the researcher can determine whether the experimental new treatment differs from the treatment as usual in a statistically and / or clinically significant way. Sometimes, people also receive one or two follow up assessments to investigate how the treatment effect in the two treatment conditions endures.
There are various ways to approach data that is collected within persons over a period of time – so called longitudinal data. Different research designs often translate to different statistical techniques. Data from simple designs as well as straightforward research questions can usually be analyzed and answered using relatively simple statistical techniques while more elaborate designs and research questions might ask for more comprehensive statistical methods. In my work as a statistical consultant, it is often a challenge to find the most suitable statistical technique, given a certain research question, design and collection of data. Sometimes, this can be a real puzzle: on occasion, with the collected data, it is almost infeasible to decently answer the research question, leaving yours truly the statistician with some sort of a casus irreducibilis. The challenge is to, even in these cases, get the most out of the data, while also being able to provide the researcher insight into the statistical voodoo you have wielded.
Besides being able to apply an arsenal of statistical techniques in clinical research, in particular techniques for the analysis of longitudinal data, I recently found a new challenge: the development and investigation of a new technique for this kind of data. This new technique, referred to as the cluster bootstrap, encompasses a relatively simple technique for more complex longitudinal data, as a contradiction to my premise above about simple statistics for simple data and complex statistics for complex data. After all, researchers often already find statistics hard to understand, so why not make life a little bit easier for them?
The development and investigation of the cluster bootstrap is the main focus of the PhD thesis I am currently working on, under supervision of prof. dr. Mark de Rooij and prof. dr. Willem Heiser. As this project develops, more information will become available on this website and on an accompanied website.