Deck 8: Model Selection in Multiple Linear Regression Analysis

Question 1

Omitted variable bias is a potential problem because it

A)prevents accurately estimating true marginal effects.
B)results in estimated standard errors that are too large.
C)results in inefficient parameter estimates.
D)might highlight spurious correlations.

Answer 1

A

Question 2

One can deal with missing data by

A)making up values for missing data.
B)creating a dummy variable for observations with non-missing values,setting the missing value equal to 1,and including the dummy variable in the regression analysis.
C)creating a dummy variable for observations with missing values,setting the missing values equal to 0,and including the dummy variable in the regression analysis.
D)creating a dummy variable for observations with missing values,setting the missing value equal to 1,and including the dummy variable in the regression analysis.

Answer 2

C

Question 3

The RESET test is used to

A)test for the inclusion of higher-order polynomials.
B)test for choosing between non-nested models.
C)test for the individual significance of coefficients.
D)test for omitted variable bias.

Answer 3

A

Question 4

If you had to either include an irrelevant variable or omit a relevant variable,you would prefer

A)including an irrelevant variable over omitting a relevant variable because omitting a relevant variable leads to biased estimates.
B)including an irrelevant variable over omitting a relevant variable because omitting a relevant variable leads to the standard errors to increase.
C)omitting a relevant variable over including an irrelevant variable because omitting a variable does not affect the estimated coefficients.
D)omitting a relevant variable over including an irrelevant variable because including an irrelevant variable does not lead to biased estimates.

Question 5

The Eye test is used to

A)test for the inclusion of higher-order polynomials.
B)test for choosing between non-nested models.
C)critically assess the regression results instead of taking all results at face value.
D)test for omitted variable bias.

Question 6

One can deal with missing data by

A)dropping observations with missing data if the missing data is random and therefore will not affect the resulting coefficient estimates.
B)making up values for missing data.
C)ignoring the potential impact of the missing data in the analysis.
D)running a regression with the number 6 put in lieu of the missing data.

Question 7

One can deal with potential outliers by

A)dropping them from the data set.
B)including a dummy variable equal to 1 if the observation is an outlier.
C)performing Weighted Least Squares.
D)dividing the value of outlier by the sample mean.

Question 8

The Davidson-MacKinnon test is used to

A)test for the inclusion of higher-order polynomials.
B)test for choosing between non-nested models.
C)test for the individual significance of coefficients.
D)test for omitted variable bias.

Question 9

Suppose that you are performing the Davidson-MacKinnon test for choosing among non-nested alternatives and that in the second stage you estimate the sample regression function and find that the predicted values are not statistically significant.You decide that

A)the initial model is statistically preferred.
B)the second model is statistically preferred.
C)neither model is appropriate.
D)the coefficient estimates from both models are biased.

Question 10

Inclusion of irrelevant variables is a potential problem because

A)results in biased estimated slope coefficients.
B)might highlight spurious correlations.
C)results in estimated standard errors that are too large.
D)prevents accurately estimating true marginal effects.

Question 11

Suppose that you estimate the sample regression function
You might be concerned that
Is a biased estimate of the marginal effect of experience on salary because

A)a person's sex is likely an irrelevant independent variable.
B)marital status is likely an irrelevant independent variable.
C)education is likely an omitted independent variable.
D)the identity of the last NCAA basketball champion is likely an omitted independent variable.

Question 12

Suppose that you estimate the sample regression function
Employing the "eye test" you might suspect that the marginal effect of

A)experience is estimated incorrectly because it does not seem reasonable that each additional year of experience increases salary by 6.2 percent,holding all other independent constant.
B)education is estimated incorrectly because it does not seem reasonable that each additional year of education increases salary by 3.8 percent,holding all other independent constant.
C)having Blue Eyes is estimated incorrectly because it does not seem reasonable that the average salary of individuals with Blue Eyes is 853.8 percent less than the average salary of individuals with other eye colors,holding all other independent constant.
D)GPA is estimated incorrectly because it does not seem reasonable that each one unit increase in GPA increases salary by 1.9 percent,holding all other independent constant.

Question 13

Outliers are potentially problematic because they

A)result in biased estimates.
B)skew the data to the right.
C)skew the data to the left.
D)result in larger estimated standard errors.

Question 14

Potential inclusion of irrelevant variables is best dealt with by

A)carefully considering economic theory.
B)performing tests of individual significance.
C)performing tests of joint significance.
D)including the fewest possible independent variables.

Question 15

All of the following are potential problems associated with missing data except

A)dropping observations with missing data will bias the estimated coefficients if the missing data is due to selection bias.
B)the number of observations goes down.
C)dropping the independent variables that have missing data will typically bias the estimated coefficients.
D)missing data always increases the estimated standard errors.

Question 16

Suppose you estimate restaurant sales as a function of the quality of SERVICE,PRICE,and consumer INCOME and you get the following results (p-values in parentheses)
Given our usual language,we would conclude that

A)SERVICE,PRICE,and INCOME are all statistically significant.
B)INCOME and SERVICE are statistically significant;PRICE is marginally significant.
C)INCOME is statistically significant;SERVICE and PRICE are marginally significant.
D)INCOME is statistically significant;SERVICE are marginally significant;PRICE is statistically insignificant.

Question 17

Potential omitted variable bias is best dealt with by

A)performing tests of individual significance.
B)carefully considering economic theory.
C)including all possible independent variables.
D)performing tests of joint significance.

Question 18

Suppose that you are performing the RESET test for the inclusion of higher-order polynomials and that in the second stage you estimate the sample regression function and the predicted value terms are statistically significant.You decide that

A)higher-order polynomials are not necessary for this regression.
B)you should investigate the inclusion of higher-order polynomials in your regression model.
C)neither regression model is appropriate.
D)higher-order polynomials are never appropriate to include in regression models.

Question 19

Omitted variable bias occurs when one does not include

A)an independent variable that is correlated with the dependent variable only.
B)an independent variable that is correlated with the dependent variable and an included independent variable.
C)an independent variable that is correlated with an included independent variable only.
D)a dependent variable that is correlated with an included independent variable.

Question 20

Suppose that you estimate the regression function
You might be concerned that
May have a large standard error because

A)investor wealth is likely an irrelevant independent variable.
B)the amount of rainfall is likely an irrelevant independent variable.
C)investor wealth is likely measured with error.
D)the first letter in the firm's name is likely an omitted variable.

Question 21

Suppose you are interested in estimating how the test scores of elementary schools are related to average class size,parents education level (in years),and percent of English learners at the school.To do so,you collect a sample of 200 California public schools and specify the following model:
Suppose you are concerned that the proposed model needs higher order polynomials.
a)Initially you believe that class size is the only variable that should be entered as a higher order polynomial.Propose a new model and explain how you would test for this possibility.
b)After thinking about it,you believe that all of the independent variables may need to be entered in as a higher order polynomial.What type of test would you perform? Describe the steps you would take to implement this test.
c)Now instead you believe that the model estimated above is not appropriate and a better model would be
How would you test between the initial model and this model? Be as specific as possible.

Question 22

Why is missing data a potential problem? What are two ways to deal with it? Which approach do you prefer and why?

Question 23

When would you use the Davidson-MacKinnon test? What is the null hypothesis for the test? What is the intuition for why it works? Explain.

Question 24

Suppose you are a potential college student that is interested in determining whether it is worthwhile to declare a certain major.In an effort to find the answer,you collect data on 1,247 college who majored in HUMANITIES,SCIENCE,ENGINEERING,or ENGLISH and you estimate the sample regression function (standard errors in parentheses)
a)Do you think omitted variable bias is a potential problem in this case? Why? Explain.
b)What is the problem associated with omitted variable bias? Explain.
c)How might you control for the potential omitted variable bias in this case? Explain.

Question 25

When would you use the RESET test? What is the null hypothesis for the test? What is the intuition for why it works? Explain.

Question 26

Suppose you are interested in the factors that explain GDP all over the world.You gather data for the most recent year on GDP,Education level,Household Consumption,and Number of People Employed for every country in the world and you specify the following model.
a)You are able to find data for GDP,Household Consumption,Number of People Employed for every country but Education Level is missing for 40% of the countries.What are your options when estimating this model?
b)What type of countries do you think have missing data for Education Level? Why?
c)If you ran a regression with only the 60% of countries that don't have missing values,do you think your results would be different than if you had been able to gather data on all countries? Why or why not?

Question 27

What is omitted variable bias? Why is it a problem? How do you try to prevent it? Explain.

Question 28

What is the "eye test"? Why is it important to employ it every time you estimates a sample regression function? Explain.

Question 29

What is the inclusion of an irrelevant variable? Why is it a problem? How do you try to prevent it? Explain.

Question 30

What is an outlier? Why are they a potential problem? What can you do to deal with outliers? Why does this approach work? Explain.

Question 31

What is the potential shortcoming of using a strict cutoff to determine statistical significance? Explain.