Partially Adaptive and Semiparametric Estimation of Regression Models for Grouped Data

• Activity Number: 584
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Jason Blaine Cook*+ and James B. McDonald
• Companies: Brigham Young University and Brigham Young University
• Keywords: Partially adaptive estimation ; Semiparametric estimation ; Grouped data ; Interval data

Abstract

Several valuable data sources, including the census and NLSY, include data based on interval responses. Many empirical studies attempt estimation by assuming the data correspond to the interval midpoint and run OLS regressions or performing maximum likelihood estimation (MLE). Stata provides a MLE approach, based on the assumption of normality, which allows for intra-group variation. In the presence of misspecification, these estimates are inconsistent. Other solutions include utilizing nonparametric estimation techniques. In this paper we take an intermediate position, using partially adaptive estimation, which builds on a MLE framework using flexible parametric probability density functions (pdfs). Semiparametric procedures are also considered. Applications of these methods are used to estimate determinants associated with household income based on US Census data. Monte Carlo Simulations are performed to evaluate the relative efficiency of the different methods of estimation.