[Data cleaning in the 1995-1996 national smoking prevalence survey]

The authors introduced the procedure of data cleaning in the 1995-1996 national smoking prevalence survey, and the objectives, principles, procedures, quality evaluation and points for attention of data cleaning were summarized. Data cleaning must be carried out from the smallest investigative unit. The logical relationship among variables must be fully considered for the revision of incorrect values. In dealing with miss values, their roles in the statistical analyses must be evaluated. Comparing the differences in some important variables before and after cleaning is a useful method for assessing whether the representativeness of the data has been changed during the cleaning process.