Analytical modeling in complex surveys of work practices

Quantitative industrial relations research frequently relies on data collected from large surveys of establishments that use complex sampling designs, such as stratified and unequal probability sampling. The authors analyze two complex surveys of establishments, the National Organizations Survey and the National Survey of Establishments. They discuss design-based (survey-weighted) and model-based (unweighted) strategies for analyzing these data. They show that the choice of strategy can affect inferences about parameters, and hence conclusions drawn from analyses. They discuss the advantages of model-based approaches that include independent variables corresponding to design features, such as functions of size measures or indicator variables for strata or clusters, relative to purely design-based approaches. © by Cornell University.

Duke Scholars

Author Jerome P. Reiter Statistical Science