A1694
Title: Estimators of variance for matching-based estimates in the setting of complex surveys
Authors: Karen Messer - University of California, San Diego (United States) [presenting]
Abstract: The large sample properties of resampling estimates of variance for matching estimators of the ATT are studied in the setting of a large-scale complex population survey. Resampling-based estimators of variance must reflect the complex hierarchical sampling design of the survey. Our working example is the Current Population Survey from the US Census Bureau, which uses published replicate weights to implement a jackknife estimate of variance, using the computational refinement of Balanced Repeated Replication, a common approach. Other approaches to these data include bootstrap-based methods, either conditional on the match or incorporating the randomness of the matching process. We give an overview of methodological issues in this setting, and present some recent work.
