Matching noise: formalization of the problem and some examples

Statistical matching attempts at producing a unique, synthetic data file, where variables observed in different sample surveys are jointly recorded. Such a file is appropriate for further statistical analysis when the joint probability distribution of the variables of interest in the population coincides with the probability distribution of the same variables in the synthetic data file, or at least when these two distributions are “close enough”. The discrepancy between these distributions is called matching noise. In this paper, statistical matching methods based on hot-deck imputation procedures are investigated as a possible cause of matching noise. Two examples when data are generated from uniform and normal distributions are discussed.