Causality
Either of two counterfactual consequences when a subject is exposed to one or another treatments
What is Potential Outcome
For assessing if donut consumption increases happiness, using a measure of "sweet tooth" as an instrument fails this important assumption
What is exclusion restriction?
Borrowing a page from Euclid's postulates, this fundamental assumption is basic to DiD
What is the parallel trends assumption?
I could use this to design an RD for the effect of drinking on car accidents in the US
What is drinking age?
This color is a theorem that helped grow the popularity of OLS
What is BLUE?
This three-letter acronym allows a visual understanding of the causal relations in a complex system
What is a DAG
Levels of rainfall are entertained as an IV for measuring levels of happiness in response to donuts. Their relationship to donut consumption is known as this.
What is the first stage?
The control group in a DiD is used to estimate this for the treated group, allowing calculation of the average treatment effect for the treated
What is counterfactual?
This type of RD is used when the arbitrary rule only partially shifts the probability of taking the treatment
What is "Fuzzy RD"?
The main situation that biases OLS is when this is correlated with the Xs
What is the error?
It is the idea that we cannot observe both what happened and what _could have_ happened for any individual
What is the fundamental problem of causal inference?
The reduced form in an IV analysis is analogous to this analysis in an RCT
What is ITT?
In a 2-group, 2-period DiD regression, this term is an estimate of the treatment effect
What is an interaction term?
This helps us figure out how far away from the cutoff we can still use the data we have for RD
What is The Bandwidth?
OLS is linear in these, not in the variables
What are the parameters (or coefficients)?
This scientific tradition implores us to only use observations and data in making inferences
What is empiricism?
In the study of the Oregon Medicaid experiment, this semi-random event was used as the instrument.
What is a lottery?
Use this to make sure the shared trends are not messing up your effect estimates
What are time fixed effects?
This foible was used in an analysis that seemed to exaggerate the effect of air pollution on mortality in China
What is "choice of polynomial"? (Or anything similar)
Any of these are candidates for the least favorite - yet commonly used - statistics in OLS
What are R-squared or p-value
This Scottish philosopher argued that causality is a mental construct we assign to events that co-occur regularly
Who is David Hume
In this type of instrument, upstream shifts are coupled with local vulnerabilities to estimate the first stage
What is a shift-share instrument?
This "small cousin" of DiD allows a partially quantitative causal assessment when only a single treated unit is available
What is synthetic control?
This was used as a cutoff to design an RD examining the effect of Medicare on hospital mortality
What is Age 65?
When this test shows equivalence between regression outputs, a random effects model can replace a fixed effects model
What is Hausman test?