Controlling Individual Information in Statistics by Coding

Abstract In connection with personal integrity and privacy protection issues in statistics it has become increasingly important to control the amount of information about individual data that is contained in summary statistics. Here different numerical codes are compared and evaluated with respect to their effect on the information in a sample mean of a coded variable. Various comparison are made between the information in the mean and the information in the frequency distribution of the variable. Limiting results are given for large sample sizes.