Statistical matching is the art of combining information from two or more datasets, collected in independent samples, but a subset of the variables is common to both samples, to produce both coherent and comprehensive information in a synthetic data file, where variables observed in different samples are jointly recorded. Modern computing can make possible, under techniques described here, an advance in the application of Statistical Matching. It is reported on a simulation that splits a single file into two pieces, statistically matched the pieces using four methodologies and compares the results to the original single file. The Conditional Independence Assumption (CIA) did not seem a valid assumption. To avoid CIA, it is suggested two methods of statistical matching where kind of auxiliary information, based on Stochastic Affiliation relationship between income and rent, and residuals are used, in conjunction with nonparametric regression and hot deck distance. Both are compared not only with the classic method based on the CIA and regression without residuals but also with each other. In a nonparametric approach, with a micro objective, an unconstrained statistical matching is applied using hot-deck nearest neighbor within classes, using the household logarithm income and logarithm rent percentile groups. We are satisfied with the performance of creating a correlation coefficient of Y and Ζ , as measured using MSE.