POL S 15 Lecture Notes - Lecture 16: Causal Inference, Confounding, Lincoln Near-Earth Asteroid Research
Document Summary
If sig/p value is less than 0. 05, then the variable"s relationship is statistically significant!! Taking on alternative explanations to account for confounding variables "this is the significance of [variable] holding constant all other factors" Extraneous variables could be intervening instead of antecedent (learned last class) Identifying an intervening variable (mechanism) strengthens the causal inference, clarifying the causal connections between variables. Helps us understand more precisely how x causes y. Here, x is an indirect rather than direct cause of y, and z becomes the direct cause. Problem: if you hold z constant, x can no longer affect y because z is what links them. Then estimate another model and include another variable. Then based on the model, we see that the introduction of a new variable as raised the correlation coefficient, suggesting that the newly introduced variable plays a role in the connection of the previous two variables.