Simulation Study to Compare Imputation at the ELI-PSU Level Versus the ITEM-AREA Level

• Activity Number: 300 - Innovations in and Applications of Imputation
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Government Statistics Section
• Author(s): Onimissi M Sheidu*
• Companies: Bureau of Labor Statistics
• Keywords: Consumer Price Index; imputation; generating missing values; missing completely at random; missing at random; not missing at random

Abstract

The data for the Consumer Price Index (CPI) calculation are regularly collected on a monthly or bimonthly basis. Sometimes, however, not all prices can be collected, which creates missing values in the database used for index calculation. In the CPI, the missing prices are imputed before index calculation. Prior to the creation of the new estimation system, CPI imputed at the Item Stratum-Index Area level. Since the deployment of the new estimation system in January 2015, imputation of missing prices has been done at the Elementary level item-Primary sampling unit (ELI-PSU) level. As a result of this transition, the Imputation Research team was asked to analyze the effect of this switch. In this paper, we conduct a simulation study that compares the imputation of missing prices at the ELI-PSU level versus at the Item Stratum-Index Area level. The results of the study show that imputing at the ELI-PSU level is better than imputing at the Item Stratum-Index Area level.