Econ 570 -- Fall 2007

Identification

Logit/Probit Models

MT 2 Answers

Midterm 2

Midterm 2 Summer 2007 and More

We reviewed the "required elements" to a t-test question that asks for details.

We elected not to repeat the Cobb-Douglas questions, which are available in a couple of other places. What follows is some extra material.

Review

Left-out variables, one more time, with feeling:

Bonus material (not on exam):

Sometimes, a regression with a lower R-squared is "better."

Proof you cannot make sense out of the t ratio with one dummy variable picked to be left out:

Q & A

Testing for constant returns to scale and technological progress for the Cobb-Douglas production function:

Elasticities:

Possible heteroskedasticity corrections:

Simultaneity

Discussion of Heteroskedasticity and Serial Correlation

We discussed these two topics, but the proctor for the exam in the 12:30 class erased much of the board before I started taking pictures.

Heteroskedasticity

Discussion of Income Project

Serially Correlated Errors (cont.)

Notes for "Income" Homework and Term Paper

The "1 page" sections should be on separate sheets of paper and might not fill the page.

There are three ways to present results: equations with t stats or se's below coefficients (as in your textbook), cut and paste from Stata, and tables. The tables option is the best for a term paper, and it especially useful for comparing a few similar regressions.

Serially Correlated Errors

If the error term is left-out variables, then it is reasonable to think that it is serially correlated much like x and y.