Simulation of Incorporating the Variance from Missing Data in Census Coverage Measurement (CCM) Variance Estimates

• Activity Number: 527
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Richard Griffin*+
• Companies: U.S. Census Bureau
• Keywords: Logistic Regression ; Imputation ; Variance Estimation

Abstract

For the Census 2000 CCM, variance due to missing data imputation was incorporated into the jackknife variance estimates by the re-calculation of the missing data adjustments, such as for missing P-sample match status, for each jackknife replicate. For each cell the weighted proportion of P-sample matches for resolved cases was used as the imputed probability of a match for unresolved cases in that cell. For the 2010 Census CCM, imputation will be done using logistic regression. A possibility for variance estimation is re-computing the estimated logistic regression coefficients for each jackknife replicate to account for missing data variance. This paper documents an empirical study using simulated data to evaluate the bias and variance properties of a jackknife variance estimate that includes variance due to a missing data imputation using logistic regression.