Sample size optimization for person- specific temporal networks using power analysis and predictive accuracy analysis

Zhang, Yong (2023) Sample size optimization for person- specific temporal networks using power analysis and predictive accuracy analysis. Research Master thesis, Research Master.

Text 20230629_Masterthesis_final_YZ.pdf Restricted to Repository staff only Download (2MB)

Abstract

Inspired by the network theory of psychopathology, many recent studies adopted experience sampling methods to collect intensive longitudinal data (ILD) on psychopathological symptoms. To capture the within-person dynamics of these symptoms, the lag-1 vector auto-regressive (VAR(1)) model is widely used and its results can be further visualized as a person-specific temporal network. Such person-specific networks are promising tools for clinical practice, generating personalized insight into the pathology and treatment of mental disorders. However, the promise is hampered by the risk of the VAR(1) estimates and the associated network being of low quality. An important determinant of the model quality is the number of time-points collected in ILD, in that this sample size determines the power of the significance tests performed on the edges of a network, as well as how well a network can generalize to unseen data from the same patient. In this study, we propose two simulation-based methods that aim to find the optimal sample size for such network analysis: power analysis and predictive accuracy analysis. Two applications of both methods are demonstrated: (1) "a-priori"—informing the sample size planning of future network studies, and (2) "post-hoc"—evaluating whether the sample size of existing network studies was large enough to ensure sufficient quality of the networks. Results suggest that commonly used sample sizes in previous single-case network studies may not be large enough, leading to high risks of underpowered edges and unsatisfactory predictive accuracy of the networks. Further implications for future network studies in clinical research and practice are discussed.

Item Type:	Thesis (Research Master)
Supervisor name:	Bringmann, L.F.
Degree programme:	Research Master
Differentiation route:	Psychometrics and Statistics [Research Master]
Date Deposited:	19 Jul 2023 08:38
Last Modified:	19 Jul 2023 08:38
URI:	http://gmwpublic.studenttheses.ub.rug.nl/id/eprint/2432

Actions (login required)

View Item