Tzu-Yao Lin

Department Methodology and Statistics, CAPHRI
Maastricht University

Can we trust our numbers? Quantification of measurement reliability for intensive longitudinal data

In many scientific domains, like medical and psychological sciences, technical advances led to the development of devices that can record biological, physical, behavioural or environmental information in real time. Recordings can be passive (e.g., sensors) or active (e.g., questionnaires sent on a smartphone). These devices offer the possibility to study individuals in real-time and real-life settings. They generate intensive longitudinal data (ILD) on one or more variables (e.g., heart rate and stress level). ILD are characterized by many observations very close in time.

Reliability and agreement studies contribute to assessing the quality of measurement instruments by providing information about the amount of error inherent to any diagnostic, score or measurement. Unfortunately, no encompassing statistical framework with clear guidelines and a user-friendly software tool exists to study the reliability and agreement for ILD.

In this doctoral project, we propose to develop general Bayesian state-space models to assess reliability and agreement for ILD. The framework will be developed for a wide range of outcomes (binary, Gaussian and bounded) and implemented in a free statistical software. We further propose to empirically validate how model assumptions affect reliability and agreement. Based on the findings, guidelines to assess reliability and agreement for ILD will be formulated. 

This doctoral project will provide a solid statistical framework with large potential impact, given the increased use of mobile technology innovations and the lack of methods to assess their reliability and agreement.

Prof. F. Tuerlinckx
Drs. S. Vanbelle
Prof. E. Ceulemans
Prof. P. Kuppens