George Mason University
CSI/Statistics Colloquium Series
Seminar Announcement


The Use of Multiple Imputation for Calibration in Public-Use Samples

Nathaniel Schenker


National Center for Health Statistics


ABSTRACT
Every decade, the responses in the United States census concerning type of employment are grouped into industry and occupation categories. The industry and occupation classification schemes are revised for each census to portray the current characteristics of the population accurately. Major revisions were made for the 1980 census, especially for occupations. The result of such revisions has been to make analyses of change in industry and occupation structures, and change in the relationship of industry and occupation to other phenomena, very difficult. In this talk, I will describe a large project in which multiple imputation, a method for handling missing data in surveys, was used to calibrate industry and occupation codes in 1970 census public-use samples to the 1980 standard. The talk will include: (i) a brief introduction to multiple imputation; (ii) a discussion of modeling and estimation issues that arose in the census project, (iii) a presentation of results from a Monte Carlo study of interval estimates involving multiply-imputed industry codes, and (iv) some simple analyses of the new public-use samples comparing multiple imputation with an alternative method.


Friday, November 19, 1999
George W. Johnson Center, Assembly Room B
Seminar at 10:45 a.m.
Refreshments at 10:30 a.m.
For the 1999 Fall Seminar Schedule, go to
http://www.science.gmu.edu/statseminars