October 9, 2018
2:00pm – 3:00pm Los Angeles
Venue
Center for Community Health, UCLA Wilshire Center
Panteha Hayati Rezvan, PhD
Postdoctoral Fellow
UCLA Global Center for Children and Families
CHIPTS
Tuesday, October 9, 2pm – 3pm
Center for Community Health, UCLA Wilshire Center
10920 Wilshire Blvd., Suite 350, Room 350-46 (Conference Room)
*** LIGHT REFRESHMENTS WILL BE SERVED ***
We will begin promptly at 2:00 p.m.
Missing data are frequently encountered in longitudinal studies with repeated follow-up waves and have the potential to induce bias and reduce precision if not handled properly. Multiple imputation (MI) has emerged as a successful framework for addressing missing data problems, but in longitudinal settings, incorporating all variables that predict other data values or that predict non-response can lead to large imputation models. Consequent convergence issues due to high correlations between variables carry the potential for over specified models, and accompanying bias and variance inflation.
The first part of this seminar will provide an overview of multiple imputation strategies with a focus on contextual factors that contribute to sensitivity in research findings. The second part of the seminar will focus on the application of multilevel multiple imputation using data collected through a HIV prevention community-clustered randomized controlled trial in South Africa. To assess the robustness of inferences to proper handling of missing data in outcomes and exposure, I will illustrate two MI strategies developed within the joint-modeling and the fully-conditional specification frameworks, and will compare the findings with the results obtained from longitudinal mixed-effects models. Findings from this study will be discussed with the aim of offering methodological insights with potential application to behavioral and biomedical research settings.
This event is hosted by the Center for HIV Identification, Prevention, and Treatment Services (CHIPTS) and made possible by funds from the National Institute of Mental Health (MH058107).