Exam 1 Trivia Flashcards
(18 cards)
Draw a fork DAG.
Z influences X and Y, X influences Y
Provide a three-variable example of a fork, explaining each relationship
DAG provided in previous question.
smoking (X ) → lung cancer (Y ), but not the other way around (no
reverse causality)
genes (Z ) → smoking (X ), but not a two-way relationship (violates
acyclic characteristic of DAG)
genes (Z ) → smoking (Y ): it is a confounding relationship
What is a mediator? Explain and draw the DAG.
X influences Y through M
The mediator is M in the above DAG. Mediation is the process by which a
third variable (i.e., the mediator) transmits the effect of an independent
variable on a dependent variable.
What is a direct effect, and how does it relate to mediation?
A direct effect refers to the mediation effect produced every other mediator except M
In the language of DAGs, what does it mean to close all relevant backdoor paths?
It means to only close the backdoor paths associated with confounders, not colliders or mediators — i.e., assuming an interest in some form of an average treatment effect as the estimand
What is a collider? Draw the DAG and explain.
X and Y influence C
Colliders (C ) are variables that, if adjusted for, can introduce a spurious
relationship between X and Y .
What is selection bias and how do you indicate it in a DAG?
Selection bias is when you have of availability of data in your sample that
does not represent the population for which you are making your inference.
You indicated selection bias with S
What is ignorability in the context of internal validity?
No unmeasured confounders
What is ignorability in the context of external validity?
No unaccounted for selection biases that generate difference between the
sample and population in terms of effect modifiers
What are generalizability and transportability?
Generalizability is when the sample is embedded within the population of
interest, and transportability is when the sample corresponds to another
population of interest.
What is attrition? Provide an example to show that you understand it.
Attrition is when units drop out of your sample. For example, if I run a field experiment that provides malaria bed nets to people, but I can’t find the person when I come back after three months to see if they used the bed net, then that would be an example of attrition.
How might we test if attrition is a problem? Be specific
We would want to run a balance test to see if our confounders (Z ) have fundamentally different values in the treatment and control groups. If so, the attrition is differential (not random) and thus problematic. We can test for the imbalance via p-values or, even better, via standardized mean differences
What is an Intent to Treat (ITT) effect?
The effect of assigning the treatment, even if people did not comply with their treatment assignment. The ITT is often a very conservative estimand
What is the main estimand that we can recover from an instrumental variables design? Explain. Don’t just provide the
name.
The Local Average Treatment Effect (LATE). It is only relevant for compliers—i.e., those who take the treatment if assigned to it.
Explain the instrument that won Acemoglu, Johnson, and Robinson (2001) the Nobel prize. Why was it a valid instrument? Hint: your answer will be more complete if you draw the DAG.
Q influences X, which influences Y, Z influences X and Y
X : protection against expropriation
Y : log GDP
Q: settler mortality at time of colonization
Z : latitude
How can we assess the external validity of instrumental variable designs?
A good first step is to look at the R2 and pairwise correlations from the first stage. If they are low, that means that the compliers—assuming that they can be conceptualized—provide a low share of the overall variation, thereby making external validity a challenge
Why is qualitative information so important for standard natural experiments?
We need qualitative information to validate that the treatment assignment is as good as random. Otherwise, it is not a natural experiment.
Draw the two potential DAG(s) for standard natural experiments and explain your rationale
Fork & Irrelevant Z: One - X influence Y and Z influences X and Y. Other - Z and X influence Y.
Sometimes there will be real random assignment that the researcher does not control (Irrelevant Z ). Other times, the as-if random assignment won’t be that credible, so we will need to control (fork DAG).
