ABSTRACT
Nonexperiments (NE) include any design that uses statistical controls. Any randomized field experiment (RFE) that uses statistical controls for pretest scores (to improve internal and/or statis tical validity) is partly a nonexperiment, though at its core remains a randomized experiment. Most of the quasi experiments (QE) that we discussed in the last chapter also use statistical con trols, although, at their core, they use elements of matching comparable groups at one point in time and/or matching the same group or groups to itself before and after a program change. In these two cases, the line between the nonexperiment and the RFE and the line between the nonexperiment and the quasi experiment is not clear. One would describe these as mixed designs: part RFE and part NE, or part QE and part NE.
