CANCEIS Experiments of Edit and Imputation
with 2006 Census Test Data
Bor-Chung Chen
KEY WORDS:
ABSTRACT
In this report, we demonstrate the CANCEIS (CANadian Census Edit and Imputation System) experiments
of edit and imputation with the 2006 test data. The major effort is to translate the if-then-else
rules of current edit and imputation system of the decennial census into the decision logic tables (DLT)
of CANCEIS. We also formulate the input files that are needed to run the CANCEIS software. The advantages
of using DLT are that it is easy to understand the edit rules; and DLTs are input, not part of
the software, making it easier to change when edit rules are changed. We also compare the imputation
results between the CANCEIS experiments and the 2006 Census Edited File. The comparison is for our
curiosity beause the constructed DLTs are not identical to the edit rules specified in the 2006 edit specs.
Although the edit rules are not identical, the comparison still shows some similarities between the CEF
and CANCEIS results.
CITATION:
