June 25, 2009
Technical Issues (mostly statistical) Associated with the ACS
SAS code for using Replicate Weights
Statistical packages such as stata can handle replicate weights to produce standard errors and tests of significance. However, for the ACS and CPS, the replicate weights are based on a successive difference model rather than a jack-knife or generalized linear model. No statistical package handles successive difference models as an option.
Thus, one has to go through the actual algorithm to produce standard errors based on the successive difference replicate weights. The following program provides the code for this:
SAS code for using replicate weights
Garrett, B. Dale and Michael Starsinic. 2008. "ACS Public Use Microdata Samples of 2005 and 2006 - How to Use Replicate Weights." Presentation at AAPOR Conference, New Orleans, May 16, 2008.
Theoretical basis for successive difference model is:
Fay, R. and Train, G. 1995 "Aspects of Survey and Model-Based Postcensal Estimation of Income and Poverty Characteristics for States and Counties," Proceedings of the Section on Government Statistics, American Statistical Association, pp. 154-159.