The aim of the project is to develop new methods to address item and unit nonresponse in longitudinal studies with a multilevel design with a special focus on the National Education Panel Survey (NEPS). Two independent approaches are proposed to tackle this problem. Item nonresponse will be addressed by new imputation routines that account for the multilevel design. Innovative weighting methods will address unit nonresponse on all levels. Longitudinal or hierarchical data structures are often ignored when missing values are imputed although it has been shown repeatedly that this can induce bias in the imputed datasets. Still, the literature on hierarchical imputation models is very limited and the models proposed so far are hardly ever applied in practice. We will evaluate and extend the existing models, and also propose an alternative imputation modeling approach that might facilitate the imputation of hierarchical data structures compared to the more complex models discussed in the literature.. To deal with unit nonresponse on different levels of a multi-stage longitudinal survey, we will develop weighting adjustments making efficient use of paradata, linked administrative data, and calibration algorithms We will evaluate the different imputation and weighting methods using extensive simulation and empirical studies. Empirical data from the NEPS surveys will be used as a testbed for this research. This project will provide technical guidance, examples, and general purpose imputation and weighting code that could be used to handle missing values (on item and unit level) in past and future NEPS waves and other educational surveys.