552 – Contributed Oral Poster Presentations: Social Statistics Section
Empirical Studies on Market Microstructure Models
Feng Liu
The University of North Carolina at Chapel Hill
Ruiwen Zhang
SAS Institute
Surveys of influenza vaccination coverage that enable production of estimates within the influenza season often involve collection of weekly survey data, typically based on relatively small sample sizes yielding relatively high variability. Such variability also adversely affects the stability of estimates across time, the result being estimated trends that may show occasional declines, even though the true population trends are by definition non-decreasing. Composite estimation, utilizing data and combining estimates across time periods, offers the opportunity for more stable estimates of coverage levels and trends as well as estimated trends less subject to period-to-period decreases. Use of survival analysis techniques is another alternative that ensures non-decreasing estimated trends. This paper profiles variability associated with direct estimates of levels and trends associated with the influenza module of the National Immunization Survey, proposes a composite estimation and a survival analysis approach for combining data across time, assesses the variability associated with composite and survival estimates of weekly influenza vaccination rates, and discusses potential error associated with use of data collected in different survey periods.