In 2017, Noémi Schuurman was appointed assistant professor in the Department of Methodology and Statistics at the Tilburg University School of Social and Behavioral Sciences. With a great amount of knowledge of and experience in analyses of individual networks and intensive longitudinal data, Noémi is an asset to the School and to Tilburg Experience Sampling Center (TESC). She mentions TESC as one of the reasons why she came to Tilburg University. “TESC was a sign to me that there is interest in my type of research at Tilburg University”. TESC talks with Noémi about the impact of her work on the research field of experience sampling and intensive longitudinal data.
About Noémi Schuurman
Noémi is a methodologist and has been focusing on the analysis of intensive longitudinal data since her studies in 2010. Her work on multilevel VAR modeling has been implemented in popular statistical software Mplus. During her internship in 2010 at the University of Amsterdam, when network analysis was still in its early stages, she worked on dynamic networks together with Denny Borsboom. She did a phd-project on multilevel time series models at Utrecht University with Ellen Hamaker as her supervisor, Herbert Hoijtink as her promotor. Currently, she works as an assistant professor at Tilburg University and has several publications in international journals, such as Psychological Methods.
How it all started
Noémi’s career started with her studies in Psychology at the University of Amsterdam. When she signed up for a Psychology study, she initially thought of becoming a therapist/practicing clinical psychologist, but she instead decided to specialize (major) in Methodology (both in her bachelor and master). Because of her interest in aiding individuals, it was a natural choice to focus on N=1 analyses and intensive longitudinal data, because these methods can be tailored to single individuals. These analyses were not yet widely used in psychology at that time. In 2010 Noémi does an internship with Denny Borsboom (known for network models in the context of psychology) and is asked to make a network based on time series, together with her fellow student Robert Hillen. “Denny told me: You need to go do something with time series! So I dove into the time series literature, and started by making an N=1 network out of autocorrelations and cross-correlations, and later with auto-regression and cross-lagged coefficients estimated with VAR(1) models.” These ‘dynamic (VAR) networks’ are now frequently used in psychology. “That’s how it all started for me.”
Using VAR models to obtain person-specific granger-causal networks (2010)
Multilevel time series analyses
After her Research Master’s in Psychology, Noémi did a PhD project in Utrecht under the supervision of Ellen Hamaker, a leading expert in the field of time series analysis. Noémi worked on a project on Bayesian multilevel (vector) autoregressive models, and she explains what this means: “Ordinary time series analyses apply to one person, but, as a psychologist, you often want to generalize to a group of people. This is difficult with N=1 analyses. The multilevel model makes it possible to analyze multiple people at the same time while still taking into account that people differ from each other”. In her work, Noémi focuses on exposing fundamental problems in methods for analyzing psychological intensive longitudinal data, providing solutions to those problems, and developing new methods that make new applications possible. Noémi is now working at Tilburg University, on a tenure track position in the Department of Methodology and Statistics of the Tilburg School of Social and Behavioral Sciences, and Noémi is specialized in the field of (Bayesian) multilevel time series analysis.
Developing new techniques
What Noémi would like to achieve is to further develop intensive longitudinal data analysis techniques that are appropriate particularly for psychology. “There is still a lot of work to be done in this area analysing intensive longitudinal data” says Noémi, “there is more work in this area in other fields, like economics, but there is not much that really corresponds to psychology yet.” She enjoys looking at statistical models and thinking about questions like: How does this model correspond to a psychological theory? How does it not correspond, and how can we do something about that? What kind of problems come up with these analyses, and can we solve this? “These questions can lead you in many directions and to many topics, but that is the luxury of being a methodologist. What I ultimately hope for is that my work can somehow be of use to individuals, for instance via clinical applications, but that is a more indirect goal.”
Working with impact
Noémi is proud but also slightly surprised that her work has impact. “That’s great, and you don’t necessarily expect that. It always is a pleasant surprise when people read your work carefully.” She is happy with the fact that a large company like Mplus wanted to implement her work in popular software. Mplus implemented her standardization method for all Bayesian multilevel models. In this way, her work has a large indirect impact. ”Everyone who uses standardized coefficients in these Mplus analyses uses my method. That’s really cool.” She is also proud that she contributed to the start with individual networks and time series. “In my internship I really borrowed existing VAR models from econometrics, and made networks out of those, but it turned out to have a lot of impact in the field,” says Noémi. Her final (unpublished) internship report can be found on Denny Borsboom’s website Psychosystems.org as one of the first items on the publication list in 2010.
Experience sampling data pioneers
Noémi expects more and more attention being paid to intensive longitudinal research. Noémi is happy with the side effects of this newly developed field, in that there is a renewed attention for measurement. Noémi explains, “In cross-sectional analyses, the idea is often that it doesn’t matter much whether you do measurements today or tomorrow. However, if you want to capture processes that take place within people over time, the measurement moments can make a lot of difference. Maybe you’re missing something if you measure at the wrong time. I notice that researchers in the ILD field pay more and more attention to these kind of measurement issues, which is really essential, and something that sometimes seems to have been lost a bit in more traditional cross-sectional studies.” Important questions are: What do you capture in the measurements with intensive longitudinal data, and what not? How do you separate the data from measurement errors? Etc. “We can borrow a lot from traditional cross-sectional techniques to answer these kind of questions, but we will also have to develop many new techniques” says Noémi. “Methodologists and empirical researchers will have to do this together. I think that we will have our hands full with this in the coming period. We will have to pioneer in this field.”
Teaching and training
One of Noémi’s goals is that these analyses techniques may eventually be used to help individuals in practice. Noémi also likes to help people via the lecture hall, as she says: “Today’s students are tomorrow’s researchers.” She likes to teach and give workshops and notices that the demand for workshops on the analysis of intensive longitudinal data is very high. In the summer of 2019, she gave a summer school in Utrecht about “Modeling the dynamics of intensive longitudinal data”, together with Ellen Hamaker (Utrecht University), Laura Bringmann (University of Groningen), Rebecca Kuiper (Utrecht University), and Oisín Ryan (Utrecht University). This multi-day course will be held in Tilburg from 6 to 9 January 2020, and it will cover those analysis techniques that can be applied to ESM data and the challenges this field of work faces.
More information about the course can be found on the registration page.
Paper on reliability at a personal level for intensive longitudinal data
Noémi collaborates with national and international fellow researchers. A recently published paper, a collaboration with Ellen Hamaker, deals with measurement error and reliability on a personal level. “In the area of intensive longitudinal data, people typically do not take into account measurement reliability and measurement error. The paper is about the consequences of disregarding measurement error in the ILD context, how to take measurement error into account, and how to obtain person-specific estimates of reliability.” Click here to read the paper, titled ‘Measurement error and person-specific reliability in multilevel autoregressive models’.
Noémi’s website provides an overview of papers and publications.