Peter Kennedy’s A Guide to Econometrics covers pretty much everything I’ve forgotten from the master’s degree in statistics I did a few (!!) years ago. And it does so with just the right combination of sophistication and simplicity.
This guide tackles topics ranging from criteria for estimators (least squares, unbiasedness, maximum likelihood, asymptotics, etc) to what do to when standard assumptions of linear regression are violated to Bayesian approaches to robust estimation.
I have just one quibble. Kennedy says, “What distinguishes an econometrician from a statistician is the former’s preoccupation with problems caused by violations of statisticians’ standard assumptions; owing to the nature of economic relationships and the lack of controlled experimentation, these assumptions are seldom met.”
I don’t agree that this is what separates econometricians from statisticans; many, perhaps most, statisticians deal with observational studies not experimental ones (especially in social science). All of us with our hands on data have to know the assumptions of our methods and know what to do when they are violated. That’s one reason this book is so useful: violation of each important assumption merits its own chapter instead of getting buried in a description of a particular method or in an afterthought section on testing assumptions.
Very cool; highly recommended.