Reduction of Survey Nonobservation Errors Through Adaptive Sampling Design

• Activity Number: 303
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Survey Research Methods Section
• Author(s): Shin-Jung Lee*+
• Companies: University of Michigan
• Keywords: Nonobservation errors ; Adaptive sampling ; Propensity score ; Survey sampling ; NHIS ; BRFSS

Abstract

Survey nonobservation errors are an increasing problem in surveys of all modes. This paper examines the nonobservation errors for survey estimates with two different sampling designs: fixed sampling design and adaptive sampling design. Both designs are assumed the same design features during data collection. The difference is in how the auxiliary variables are used. Fixed sampling design uses auxiliary information through post-survey adjustments to eliminate nonobservation errors. The adaptive sampling design, utilizing those auxiliary information at the design stage to make sampling decision, reduces nonobservation errors at each data collection phase. The nonobservation errors are studied by simulated hypothetical surveys under comparable auxiliary information and nonresponse mechanisms. The simulation studies are based on real data from two large government surveys, NHIS and BRFSS. The nonobservation errors, measured by biases and variances, for estimated means and ratios are investigated for the full population and for domains. The results suggest the benefits of adaptive sampling design in reducing nonobservation errors.