Assumptions

While estimates can be calculated with any sort of data, we need assumptions to interpret the results correctly.  For example, the estimated coefficient are only unbiased under certain assumptions.

Goals:  Unbiasedness, Efficiency.

Lots of cases.  Pick a common one.

Right Model

Regressors:  Fixed in Repeated Samples  Exogenous (see also Endogenous)

Errors:  Serial Independence

Errors:  Homoskedasticity (see also Heteroskedasticity)

more