Category: Research Methods and Statistics
Keywords: Research Methods | Statistics | Longitudinal
Presentation Type: Symposium
Electronic devices such as smartphones, smartwatches, and other ‘wearable tech’ offer the possibility of continuous/repeated data collection. Such data have been used to further our understanding of important clinical topics such as the vulnerability for and onset of psychopathology, the prediction of relapse, and relapse prevention. Often, emotion states, daily experiences, or symptoms are collected several times per day for several weeks, and the goal is to infer the statistical dynamics of these items over time. The most common statistical model for such data is the vector autoregressive (VAR) model and variants thereof. In this talk, I will introduce VAR models for n=1 and n>1 data, and show how such models can be estimated using the packages graphicalVAR and mlVAR in the free statistical programming language R. VAR models provide numerous ways of investigating the data on an idiographic and nomothetic level: we can estimate how items influence each other across time (lagged effects) and within time (contemporaneous effects), can look at the networks of individuals and groups of individuals, and can explore how strongly participants differ from each other in their networks. Additionally, I will discuss prior research that used this methodology to successfully predict future transitions of patients into depression by studying so-called “early warning signals”, and cover some extensions of the VAR model such as regularization or multilevel procedures. VAR models offer crucial insights to studying both idiographic and nomothetic aspects of psychopathology, and will likely play an increasingly important role in the future of clinical methodology.
University of Amsterdam
Saturday, November 18
1:45 PM – 3:15 PM
Sunday, November 19
8:30 AM – 10:00 AM
Sunday, November 19
10:15 AM – 11:45 AM
The asset you are trying to access is locked. Please enter your access key to unlock.