POLS 208 Lecture 13: Observational Studies

• School: Emory University
• Department: Political Science
• Course: POLS 208
• Professor: Lauren Mattioli

Document Summary

Observational studies: no attempt is made to manipulate the world is reported on, gold standard for description, observe correlation, infer cause, but (cid:449)e ca(cid:374)"t sho(cid:449) it, becoming less and less common. Observational data: observational data, occurs naturally, often pre-dates the study, usually refers to large-n analysis (but not necessarily so, statistical techniques favor large-n. In the most typical case, observational data is summarized in some way, but often, that leaves us aski(cid:374)g, (cid:862)what ca(cid:374) (cid:449)e i(cid:374)fer fro(cid:373) this? (cid:863) Improving the internal validity of observational studies: matching & case-control, well-specified statistical models, control variables, appropriate functional form, accounting for error dependence. Individuals can be countries, individuals, etc: useful for outcomes that occur a long time after the cause or are rare. Control variables: helps to (cid:373)eet (cid:374)ecessit(cid:455) si(cid:373)ilar group co(cid:373)pariso(cid:374) (cid:894)(cid:862)apples to apples(cid:863)(cid:895, death penalty and crime example. Moderators and mediators: moderator: when/where an effect is observed.