Many policy interventions intended to benefit public health can only be evaluated as so-called natural experiments, because implementation is not controlled by researchers seeking to assess effectiveness. Such assessments can be complicated by non-comparability between people affected and not affected by the intervention. Various quasi-experimental designs have been proposed to address this problem of non-comparability, one being the regression discontinuity design, which has had little use in public health. This design has application when treatment assignment depends on the value of a variable—referred to as the assignment variable—reaching a threshold.

Original Source