Kalman filter models for ecological momentary assessment designs
Open Access
Author:
Lo, Lawrence Laohaopheng
Graduate Program:
Human Development and Family Studies
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
June 17, 2014
Committee Members:
Peter Cm Molenaar, Thesis Advisor/Co-Advisor Michael J Rovine, Thesis Advisor/Co-Advisor
Keywords:
Time series analysis Kalman filter Ecological momentary assessment
Abstract:
Ecological momentary assessment (EMA) designs attempt to assess individuals in real-time within real-world settings in order to provide a stronger sense of ecological validity. EMA uses temporal sampling schemes that either occur randomly or follow targeted behaviors of interest. Both of these temporal schemes produce data which contain unequal temporal spacing, violating the assumptions of many longitudinal modeling techniques. The current study examines the hybrid Kalman filter (HKF), an extension of the Kalman filter that may be particularly suitable for EMA designs. The HKF is compared with the standard discrete Kalman filter (DKF) approach in two simulation studies reflecting the different temporal schemes. The HKF demonstrated superior performance in these simulation studies; however, the DKF was robust to violations of temporal spacing and exhibited satisfactory performance in certain conditions. Finally, an implementation of the HKF with real data is demonstrated in order to provide an example of the types of results and inferences that the HKF can yield.
