Empirical selection of matching factors in matched-pairs and matched-blocks small-sample research designs

The problems of selection and proper use of factors on which treatment and control groups are to be matched are a matter of concern in all research design to control bias and increase power. In large-sample design, one does not lose by using even nonoptimal matching strategies. In small-sample design, use of factors not highly associated with outcome yields little control of bias, increases difficulties of implementing research, and can reduce power. These points are discussed and empirical strategies presented for aiding the researcher in selection and use of matching factors in small-sample research.