Douglas Rivers

The Inaugural Ross-Royall Symposium:
From Individuals to Populations
Wood Basic Sciences Auditorium
Friday, February 26, 2016
8:30 am – 6:00 pm
Douglas Rivers, B.A., Ph.D.,
Professor, Department of Political Science, Stanford University
Title:
Using Hierarchical Models to Calibrate Selection Bias
Abstract:
Samples from surveys with low response rates or opt-in samples are subject to
selection bias. Standard confidence intervals (or the margin of error) ignore
selection bias and therefore will not have the stated level of coverage for
population parameters. Using parallel surveys, we fit a hierarchical model to
estimate mean squared error, which includes both squared bias and sampling
variance. This techique provides a principled way to do statistical inferences for
internet panels without assuming ignorable selection.