Handling Missing Data in the Modeling of Intensive Longitudinal Data
Open Access
- Author:
- Ji, Linying
- Graduate Program:
- Human Development and Family Studies
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- April 04, 2016
- Committee Members:
- Sy Miin Chow, Thesis Advisor/Co-Advisor
Zita Oravecz, Thesis Advisor/Co-Advisor - Keywords:
- intensive longitudinal data
missing data
multiple imputation
full-information maximum likelihood - Abstract:
- The availability of intensive longitudinal data has helped spur the use more sophisticated methods for studying change. Unfortunately, missing data issues also arise frequently in such studies. Conventional missing data approaches are fraught with additional computational challenges when applied to intensive longitudinal data, and may not always be applicable due to the broad-ranging measurement characteristics of the covariates. In this study, we consider and illustrate the use of two approaches for implementing multiple imputing (MI) to cope with the missingness in fitting multivariate time series models, including a full MI approach, in which all missing cases are imputed simultaneously, and a partial MI approach, in which missing covariates are imputed with multiple imputation, while missingness in dependent variables are handled with full information maximum likelihood estimation. The performance of these approaches was examined under assumptions of missing completely at random, missing at random, and nonignorable missingness. The advantages and limitations of each approach are evaluated using a simulation study. We further demonstrate the implementation of the procedure in R using empirical data, involving n=111 families in which children’s influences on parental conflicts are modeled as covariates over the course of 15 days.