A1136
Title: Integrating probability and non-probability samples with misclassified covariates
Authors: Hua Shen - University of Calgary (Canada) [presenting]
Abstract: In statistics, probability samples often lack crucial, precisely measured outcome variables, whereas non-probability samples provide detailed data but are not representative. Both sample types can also involve covariates subject to misclassification. A novel two-stage estimation technique is introduced to integrate these sample types, addressing misclassification in categorical covariates using a latent-variable framework without requiring validation data. Simulations demonstrate that the method achieves superior accuracy compared to traditional approaches. Furthermore, its effectiveness is validated on a real dataset, enhancing inference quality in studies with misclassified covariates.
