The statistical technique, "Zero-One Linear Programming," that has successfully been used to create multiple tests with similar characteristics (e.g., item difficulties, test information and test specifications) in the area of educational measurement, was deemed to be a suitable method for creating multiple sets of matched samples to be used as control groups in the quasi-experimental design of "non-randomized comparison group pretest-posttest." Compared to the existing propensity-score matching method, this method does not require any statistical models and assumptions and can handle the covariate of the pretest score more appropriately. If the measurement error of the pretest-score mean of the treatment group is ignored, this method will generate a unique matched sample once the criteria for attempting to create two similar groups are determined. Otherwise, multiple sets of similar matched samples can be generated and the performance of the treatment group can be compared with each of the multiple matched samples using an appropriate statistical analysis. Afterwards, the mean of the effect size measure, taking the average of the effect size across replicated comparisons, can then be used to assess the efficacy of any program. This enhances our confidence level to decide whether a program is effective or not, compared to the finding resulting from a single comparison. A description of "Zero-One Linear Programming" and its application to create a matched sample or multiple sets of matched samples is introduced in this paper. (Contains 3 tables and 3 figures.)