Experimental Political Science

Correlation

Two variables are “correlated” when changes in one variable occur together with changes in the other

• Correlation is no causation

Spurious Correlations

When a 3rd variable (a confound) drives the correlation between two variables

Problem with correlation

All of statistics relies on correlations and there are no special “causal” estimator

Internal Validity

How confident we are that the cause (IV) really produces the effect (DV) in the study.

Confounds

An outside factor that messes up the relationship between the independent variable (IV) and dependent variable (DV).

Temporal ordering

The order you ask questions and how that might effect the answers you get

Counterfactual

the “what if” scenario of what would have happened if the cause hadn’t occured

Intervening Variables

a middle step between IV and DV (cause and effect)

Exogenous Variables

Variables that come from outside the model

Endogenous Variables

Variables that are influenced by other variables in the model

Reverse Causality

cause and effect are flipped You think x causes y but causes x

Omitted Variable Bias

Leaving out a variable that explains the relationship

Selection Bias

Results are distorted because the sample isn’t representative

False Positive

are dangerous because they give “false confidence” in something untrue.

False Negative

has an effect but data says it has no effect