Confronting Data Challenges in Research on Early Conditions and Outcome Trajectories: Methods of Combining Samples of Panel Data

Lingxin Hao , Johns Hopkins University
Stephanie D'Souza, Johns Hopkins University

Researchers often face data challenges of panel data that lacks early key information for analyzing the later panel data on outcomes. For example, estimating the impacts of early home environment on elementary school progress is impossible due to such a missing data problem. This paper extends the cross-survey multiple imputation (CSMI) method to ECSMI. The paper develops statistical rationale as well as detailed procedures in an empirical illustration of ECSMI. It borrows the joint distribution of variables in a full model from a donor sample (CNLSY) to impute two early home environment missing by design in a target sample (ECLS-K). The empirical application shows multiple utilities of ECSMI, which is to be further developed for broader application in a next step.

See paper

 Presented in Session 69. Using Linked Data Sources