Right Model

Right regressors

Right functional form

Estimation
Least Squares
Maximum Likelihood
Instrumental Variables

Conclusions
Expectation
Variance

Dummy Variables & Subsamples

Pitfalls

Why is using race (or marital status) as a regressor a really bad idea?
What is wrong with highest grade attained (or years in school) as a measure of education?
Should I worry about really small, but significant coefficients?