Matching‐Models and Examples

Abstract Evaluation is generally inherently comparative. For example, a treatment is generally neither ‘good’ nor ‘bad’ in absolute terms, but rather better than or worse than another treatment. The studies that allow for such comparative evaluations must then also be comparative. To isolate the effects of the treatments under study, avoid confounding treatment effects with unit effects, and possibly increase the power of the study, the units under study must be carefully matched across treatment groups. There are many matching techniques in practice, including crossover designs and case‐control studies. While randomization is often considered to be a competitor to matching, it actually may be considered to be a type of matching, except that here the matching is group‐to‐group.