Session 6: Large-N Part II Flashcards
What is large-N research?
Large-N research is a comparative approach where evidence across multiple cases (cross-case) is used to evaluate a causal hypothesis.
How is the counterfactual approximated in large-N research?
The counterfactual is approximated by cross-case data (what has happened to some of the units) and assumptions (made in the research design).
What do different strategies of large-N research rely on?
Different strategies of large-N research rely on making different assumptions about whether conditions are satisfied for causation to imply an association.
What are some strategies for causal inference with large-N data?
Natural experiments
Conditioning strategies
Mediating
Instrumental variables
What are conditioning strategies in large-N data research?
Conditioning strategies involve blocking (keeping constant) or measuring and including in the analysis measures of the confounders and adjusting for them.
What does mediating mean in large-N data research?
Mediating focuses on the mechanism or process that underlies the relationship between the variables in the study.
What are instrumental variables in large-N data research?
Instrumental variables refer to finding exogenous variation in the explanatory variable (the treatment) that is unrelated to the outcome.
What do all strategies for causal inference with large-N data rely on?
All strategies for causal inference with large-N data rely on statistics to rule out the influence of chance.
What are natural experiments?
In natural experiments, random assignment is produced by nature, not by the researcher.
What is regression discontinuity, a strategy related to natural experiments?
Regression discontinuity is a strategy where a (quasi-random) threshold is found that sorts cases into two groups. The variable of interest is then examined in these two groups.
What is a disadvantage of natural experiments?
The main disadvantage of natural experiments is that such mechanisms are rare.
What are the criteria for choosing an instrumental variable (IV) in research?
An Instrumental variable (IV) should meet the following conditions:
The IV is related to the main explanatory variable (MEV).
The IV is not related to the outcome variable (OV), except via the MEV.
There is no common cause Z of both the IV and OV that is not controlled for, e.g., via conditioning.
How is the effect of the main explanatory variable established using instrumental variables?
The effect of the main explanatory variable is established via a two-stage analysis:
First stage: Determine the effect of the instrument on the explanatory variable.
Second stage: Determine the effect of the explanatory variable (from residuals of the first stage) on the outcome variable.
