ANALYSES OF PUBLIC USE DECENNIAL CENSUS-DATA WITH MULTIPLY IMPUTED INDUSTRY AND OCCUPATION CODES

This paper gives a brief introduction to multiple imputation for handl ing non-response in surveys. We then describe a recently completed pro ject in which multiple imputation was used to recalibrate industry and occupation codes in 1970 US census public use samples to the 1980 sta ndard. Using analyses of data from the project, we examine the utility of analysing a large data set having imputed values compared with ana lysing a small data set having true values, and we provide examples of the amount by which variability is underestimated by using just one i mputation rather than multiple imputations.