Humboldt-Universität zu Berlin - Department of Social Sciences - Berlin Graduate School of Social Science

Cohort

BGSS Generation 2024

Title

Methods for handling informative missingness in survey data

Supervisor

Prof. Dr. habil. Sabine Zinn, Deputy Director of Socio-Economic Panel (SOEP), Head of the Division Survey Methodology and Management, Deutsches Institut für Wirtschaftsforschung e.V. (DIW Berlin)

Abstract

Missing data is pervasive in nearly all data sets. Sound statistical techniques for handling missingness are important to employ: if missing data is mishandled, invalid inference and biased results can ensue. My research topic is methods for handling informative missingness in longitudinal survey settings: when missingness means something. For a cumulative thesis, I have two areas in focus. One is the development of a new method to incorporate external data into a multiple imputation and estimate uncertainty about the method with a simulation and an application to the German Socio-Economic Panel (SOEP) study. The second research focus is censoring in time-to-event analysis: when users drop out or withdraw from the survey. I will empirically analyze whether the SOEP data set is subject to informative censoring, through a comparison with a linked pension data set. The paper will explore to what extent informative censoring influences summary statistics and report on findings. A separate paper will review different methods for coping with informative censoring that can be applied to survey data, applying those methods and comparing their effectiveness as applied to the SOEP.

Institution

German Institute for Economic Research (DIW)