ST03
Variance Estimates for Census 2000 using SAS/IML(R) Software
Contributed
Peter Davis
US Census Bureau
Abstract:
Large variance-covariance matrices are not uncommon in statistical dataanalysis. For Census 2000, we produced coverage estimates as part of theAccuracy and Coverage Evaluation (A.C.E.) Survey for 448 populationsubgroups called post-strata. Consequently, th
e A.C.E. variance estimationoperation produced a 448 by 448 variance-covariance matrix for thesenational post-strata. In obtaining variance estimates of these populationsubgroups for analysis of the census, publication, and for future research,basic ideas
of matrix theory and manipulation were essential in calculatingaccurate measures of reliability. SAS/IMLŽ software allowed for quick andeasy matrix multiplication and matrix operations to acquire the desiredvariance estimates for any combination of the 448
post-strata.
Biography:
I graduated from the University of Georgia in 1998 with a MS in statistics. I am employed at the US Census Bureau as a MathematicalStatistician. Since the Fall of 1998, I've been working on the Accurracy and Coverage Evaluation (A.C.E.) Survey for Census 2
000. I've been using SAS on a regular basis for about three years.