Ordinary Least Squares

Question 1

when is omitted variable bias not a problem?

Answer 1

collecting information on X
- better data collection reduces OVB
- concern that we can’t measure all the Xs relevant to Y

OV doesn’t affect outcome
- if X isn’t related to Y, no real factor that affects the outcome

OV is not correlated with the explanatory variable of interest (treatment)
- random assignment means that X doesn’t depend on whether you get treatment or not

Question 2

external validity vs. internal validity

Answer 2

internally valid results
- recovering treatment effects in a particular disease environment

externally valid results
- can this be applied to other populations and environments?