Sample data sourced from surveys with high nonresponse rates or ‘big data’ present challenges for practitioners using these data for inference due to their non-probabilistic traits. To address these challenges, two model-based methods enabling inference from non-probability samples include quasi-randomization and super-population modeling. The purpose of this research is to simulate non-probability samples from a synthetic population and assess the quality of the inferences derived using the proposed model-based methods. Specifically, we will apply these methods to derive estimates for a target survey measure of unemployment in California. A new approach using a linear mixed-effects model of unemployment is introduced in an application of the super-population model-based method.
