Analysis of dual system estimation in the 2000 decennial census

From a statistical point of view, the U.S. census is the most extensive sampling project. The accuracy of the census is of great interest to researchers, regulators, and the government. The Accuracy and Coverage Evaluation (A.C.E.) study was conducted to measure the overall and differential coverage of the U.S. population in census 2000. The A.C.E. involves two parts: the E-sample and the P-sample. A dual system estimation is used to obtain an estimate for the undercount for putatively homogenous groups called post-strata. A synthetic dual system estimate is finally produced for specific geographic regions. In this dissertation, we evaluate various aspects of the dual system methodology. There are two main assumptions in this process. First, the E-sample results and the P-sample results are statistically independent, i.e., there is no correlation bias. Second, the undercount rates are constant within post-strata across geographic areas, i.e., post-strata are actually homogeneous. In this dissertation, we develop a joint Poisson model to test these hypotheses. By means of a procedure involving bootstrap resampling of the log-likelihood ratio statistic and the EM algorithm, both hypotheses are rejected