Estimating Change for a Dynamic Target Population

• Activity Number: 226
• Type: Topic Contributed
• Date/Time: Monday, August 1, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Ivan A. Carrillo-Garcia*+
• Companies: National Institute of Statistical Sciences
• Keywords: Longitudinal surveys ; Panel surveys ; Weighting ; Finite population ; Super-population ; Division of Science Resources Statistics, NSF

Abstract

When the finite population of interest is dynamic, a common way to estimate change is to fix a target population as that at one fixed time, often the first wave. However, if the population steadily accrues new individuals, this requires either disregarding some data or using fluctuating weights. If individuals also leave the actual population, complexities increase. An important case is the Survey of Doctorate Recipients, a panel survey conducted by the NSF, that collects data on doctoral holders in science, engineering, or health. The goal is to estimate socioeconomic changes of the doctorate holders and to assess these across fifteen years. The major difficulties are that the target population has changed during the period of interest, with the addition of new degree holders and with loss or deletion of other subjects. Hence the usual setting for finite population inference does not occur; rather there are a set of different (albeit overlapping) finite populations through time. A superpopulation formulation addresses these difficulties. A hypothetical model generates the different finite populations; and the quantities of interest are the parameters of this superpopulation model.