Composite Estimates from Incomplete and Complete Frames for Minimum-Mse Estimation in a Rare Population

An Application Fo Families with Young Children

Published in: Public Opinion Quarterly, v. 73, no. 4, Winter 2009, p. 761-784

Posted on RAND.org on January 01, 2009

by Bonnie Ghosh-Dastidar, Marc N. Elliott, Amelia Haviland, Lynn A. Karoly

Read More

Access further information on this document at Public Opinion Quarterly

This article was published outside of RAND. The full text of the article can be found at the link above.

Random digit dialing (RDD) can be costly for a rare population, but inexpensive convenience samples are unrepresentative by themselves. The authors combine biased estimates from an incomplete frame (a listed sample) with RDD estimates in a way that improves the accuracy (Mean Squared Error, MSE) of the RDD estimates compared to what would have been achieved without the incomplete frame data. Elliott and Haviland (2007) discuss this estimator when the bias of the incomplete frame estimator is known and discuss uncertainty in estimating bias; we describe an application that estimates incomplete frame bias relative to the RDD estimate for each parameter of interest, and conditions on that estimate. The authors discuss the extent to which this approach improves MSE relative to RDD alone and relative to a common alternative-stratified estimation based on whether a case appears in the incomplete frame. They surveyed 1,002 RDD and 1,023 listed households and examined the impact of incorporating listed estimates on MSE. Conditional on the bias estimate, MSE improved substantially for many outcomes because the estimated bias of listed sample estimates relative to RDD was small for most outcomes. For thirty-eight of forty-one estimates, including the listed sample (doubling the nominal sample size) produced MSEs equivalent to RDD sample sizes 1.22-1.85 times as large as the actual RDD sample size. Because the cost per listed complete was 20 percent of the cost per RDD complete, cost per effective sample size decreased relative to RDD alone for all but three estimates.

This report is part of the RAND Corporation External publication series. Many RAND studies are published in peer-reviewed scholarly journals, as chapters in commercial books, or as documents published by other organizations.

Our mission to help improve policy and decisionmaking through research and analysis is enabled through our core values of quality and objectivity and our unwavering commitment to the highest level of integrity and ethical behavior. To help ensure our research and analysis are rigorous, objective, and nonpartisan, we subject our research publications to a robust and exacting quality-assurance process; avoid both the appearance and reality of financial and other conflicts of interest through staff training, project screening, and a policy of mandatory disclosure; and pursue transparency in our research engagements through our commitment to the open publication of our research findings and recommendations, disclosure of the source of funding of published research, and policies to ensure intellectual independence. For more information, visit www.rand.org/about/principles.

The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.