A Two-Step Approach to the Study of Changes Over Time in Right-Censored and Grouped Count Data

• Activity Number: 39
• Type: Contributed
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Kenneth Land*+ and Qiang Fu
• Companies: Duke University and Duke University
• Keywords: Right-Censored and Grouped Event Count Data, Poisson-family distributions, Intrinsic Estimator

Abstract

Response categories for frequency/count questions in repeated cross-section sample surveys often are grouped and right-censored. This limits the extent to which such data can be analyzed, as no statistical methods are readily available for modeling both the underlying frequency distributions and studying their temporal changes along age, time period, and cohort dimensions. This paper presents a two-step method that addresses problems using a minimum absolute deviations estimator to fit Poisson-family distributions to the data combined with the application of the Intrinsic Estimator for age-period-cohort analysis to study changes in the count frequency distributions over time.