POL SCI 3 Lecture Notes - Lecture 8: Internal Validity, Comparative Advantage
Document Summary
When you cannot find a match between the 2 groups. Potential confounding variables because matching only on certain variables. Cepp pre and post test, no randomization. Sol 4 build in replication, rule out fluke. Can also measure the effect of the pretest. Can look at change from pretest to posttest: if you don"t take into account preexisting values, you many underestimate or overestimate the effect. Only design with external validity, because of the comparative advantage between the crepo and crepp designs. Allows all groups to receive the treatment ethical. Matching: artificially select the group to match the other group. Blocking variable: variables we think might be possible zs (confounders), not the xs we are trying to show that have an effect on y. If we don"t block, how good is the expectation of equivalence in terms of nj. Nj: number of subjects per experimental group.