Soc 593 Statistics II
Spring 2001

Assignment 6 :  Regression Diagnostics -- Multicollinearity
Due: Wed. March 7, 2001

Part I  True/False Statements

1. Near-extreme multicollinearity does not violate OLS regression assumptions.

2. If an independent variable is highly collinear with other variables, the standard error of its coefficient will be inflated.

3. Multicollinearity does not affect the significance of the regression coefficients.

4. High multicollinearity biases the coefficient estimators.
 

Part II  Detecting and Dealing with Multicollinearity

Data: K:\nd.edu\user22\yli\Public\593sp01\Data\Hamilton\states90.dta
Task: Examining the determinants of state income level.
In this exercise, your dependent variable is INCOME (median household income in the state).  You are given the following models:
Model 1: Independent Variables: CSAT, VSAT, MSAT

Model 2: Independent Variables: METRO, DENSITY, POP, AREA

Model 3: Independent Variables: PERCENT, EXPENSE, HIGH, COLLEGE

Model 4: Independent Variables: METRO, DENSITY, PERCENT, HIGH, MSAT

Model 5: Your own specification.

Diagnose the above models and answer the following questions:
1) Is there problem of multicollinearity in each model?

2) If there is, what kind of multicollinearity is it? -- i.e., is it "Perfect Multicollinearity" or "Less-Extreme-Than-Perfect Multicollinearity"?

3) Which variable(s) are causing multicollinearity?

4) In what ways and to what extend does the multicollinearity problem affect your model estimates--i.e., does it make the coefficients biased? Inefficient? Unstable?

5) What would be the best solution to each case of multicollinearity? Justify your solutions.