POLI 210 Lecture Notes - Lecture 4: Causal Inference, Confidence Interval, Dependent And Independent Variables

• School: McGill University
• Department: Political Science
• Course: POLI 210
• Professor: Dietlind Stolle

Document Summary

It"s eas(cid:455) to (cid:272)he(cid:272)k (cid:449)hethe(cid:396) x a(cid:374)d y (cid:373)o(cid:448)e togethe(cid:396) It"s (cid:449)a(cid:455) ha(cid:396)de(cid:396) to te(cid:454)t (cid:449)hethe(cid:396) x has a (cid:272)ausal effe(cid:272)t o(cid:374) y. If estimate is precise, confidence interval is small. Causal effect is the difference in potential outcome due to treatment. Fu(cid:374)da(cid:373)e(cid:374)tal p(cid:396)o(cid:271)le(cid:373) (cid:449)ith (cid:272)ausal i(cid:374)fe(cid:396)e(cid:374)(cid:272)e is that (cid:449)e do(cid:374)"t k(cid:374)o(cid:449) the (cid:272)ou(cid:374)te(cid:396)fa(cid:272)tual. How do we ge the causal effect of treatment for a random individual in the population: average treatment effect. If the treated individuals are affected by the treatment as much as the average individual. Individuals receiving treatment are not special in any way. Interested in testing for effectiveness of a given treatment. If we just compare treated vs non-treated individuals, we are typically not going to get the causal effect right. Individuals self-select into treatment, or treatment is offered to a selected sample of individuals. Whenever you read that x causes y, ask yourself.