Topic 4: Carrying Out and Empirical Project

Question 1

If you run a regression on two variables, and the associated coefficient is negative, how are the two variables related?

Answer 1

Negatively correlated.

Question 2

When you leave a variable out of the regression and it is negatively correlated with an included variable, and the excluded variable is positive when included, you have what kind of bias?

Answer 2

Negative bias.

Question 3

To determine the bias of excluding a variable, how would you proceed?

Answer 3

Run the regression without the variable, then again with, then run the correlation of the excluded variable and the other x. The sign of correlation, along with the sign of the coefficient of the excluded variable in (in the regression where it is included) will tell you if the bias is negative or positive. Positive if the signs match, negative if they do not.

Question 4

If you remove negative bias, how does it affect the coefficient estimate?

Answer 4

It pushes the estimate up.

Question 5

If you remove positive bias, how does it affect the coefficient estimate?

Answer 5

It pushes the estimate down.

Question 6

Should you exclude highly correlated variables from a regression due to the No Perfect Collinearity Assumption?

Answer 6

You can, as long as they are not perfectly correlated. This cans sometimes be good, as long as each variable brings something unique to the regression.