Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 75 - Contributed Poster Presentations: Biometrics Section
Type: Contributed
Date/Time: Monday, August 3, 2020 : 10:00 AM to 2:00 PM
Sponsor: Biometrics Section
Abstract #313693
Title: A State-Space Approach for Analyzing Longitudinal Outcomes
Author(s): Alicia Chua* and Yorghos Tripodis
Companies: Boston University School of Public Health and Boston University School of Public Health
Keywords: Kalman filter; local linear trend; state-space model; linear mixed-effects model; random effects model; longitudinal data

Longitudinal outcomes are important to assess the disease state of a patient. The distribution of these outcomes is often skewed and does not exhibit a linear trajectory. We introduced the adjusted local linear trend model to handle these challenges. The model involves two equations - measurement and state. Covariates of interest are estimated via the state equation with the flexibility to adjust for the time elapsed between visits via a transition matrix and interaction terms between covariates. This model has a maximum likelihood estimation step for the unknown variances prior to feeding the input values into the recursive Kalman Filter and Kalman Smoother algorithms to obtain unbiased model estimates. When data from 164 subjects who completed the Boston Naming Test were used in our simulation study, our proposed model appeared to perform just as well as the linear mixed-effects models within the equally spaced-time intervals analysis but the model was superior to the mixed-effects models in the unequally spaced-time intervals analysis. Our proposed model was able to attain the lowest variance for the estimates versus other models compared with up to 20% variance reduction.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program