Chapter 5 (5-5.3.1) - Estimating Causal Effects With Observational Data Flashcards
To estimate causal effects using observational data, you first have to do what?
Identify any relevant differences between the treatment group and the control group.
What’s a confounder and how is it denoted?
A confounder is a type of variable that affects 1. the probability of receiving the treatment variable X and 2. the outcome variable Y. It’s denoted as “Z”.
True or false: confounders obscure the causal relationship between X and Y.
True
True or false: when there is a confounder affecting the treatment variable and the outcome variable, you should still trust correlation as a measure of causation.
False
In the presence of confounders, does correlation automatically equal causation?
No.
Randomization of treatment assignments eliminates all what?
All confounders.
Simple linear models use what to predict Y?
Only one X variable.
True or false: to measure causal effects, you need to compare the factual outcome with the counterfactual outcome.
True
What’s the fundamental issue of measuring causal effects?
You can never observe the counterfactual outcome.
To estimate causal effects, you must what?
Find or create a situation in which the treatment group and the control group are comparable.
True or false: in randomized experiments, you can’t rely on random treatment assignments.
False
What could a three variables-situation/-dataset look like?
The relationship between the type of school/education a students attends, how well they did on an exam, and whether or not they received private tutoring.
True or false: you can’t rely on random treatment assignment to eliminate potential confounders.
True
True or false: when X is the treatment variable and Y is the outcome variable, the estimated slope coefficient is equivalent to the difference-in-means estimator.
True
What’s observational data?
Data collected from naturally occurring events.
Should there be confounders if the data is collected from a randomized experiment?
No.
