Deck 5: Multivariate Ols: Where the Action Is

Question 1

Adding more independent variables into the model necessarily reduces bias.

Answer 1

False

Question 2

Measurement error in the dependent variable causes our beta-hat estimates to be biased.

Answer 2

False

Question 3

By adding more independent variables into our OLS model, we have a greater chance of getting rid of the endogeneity that exists within the error term.

Answer 3

True

Question 4

Adding more control variables will always increase the R² value.

Question 5

We necessarily do not have an omitted variable bias problem the omitted variable is uncorrelated with the included variable.

Question 6

Adding more independent variables can reduce multicollinearity.

Question 7

Perfect multicollinearity means all independent variables are uncorrelated with each other.

Question 8

An omitted variable bias problem occurs when:

A) We fail to include an independent variable that is not correlated with the dependent variable but is correlated with the main independent variable of interest.
B) We fail to include an independent variable that is correlated with the main independent variable of interest and the error term.
C) We fail to include the main independent variable of interest in the model.
D) We fail to include an independent variable that is correlated with the error term but is not correlated with the main independent variable of interest.

Question 9

When using an auxiliary equation to help us think though an omitted variable bias question, if $\delta$ ₁ is equal to zero then:

A) We have an OMV bias problem and therefore need to include the variable (if we can).
B) The omitted variable is not correlated with the main independent variable of interest, and we therefore do not have an omitted variable bias problem by omitting that variable.
C) We don't care about $\delta$ ₁ in the auxiliary equation and instead need to focus on the size of the error term in the auxiliary equation.
D) Need more information to answer this question

Question 10

Which of the following are consequences of measurement error in the dependent variable?

A) Smaller R²
B) Biased coefficient estimates
C) The bigger the measurement error, the bigger the variance of the error term.
D) Both A and C

Question 11

If the measurement error is in the independent variable, then

A) We don't need to worry about bias, the measurement error will be reflected in the error term.
B) The bigger the measurement error, the bigger the variance of the error term.
C) We will have a case of attenuation bias, where the coefficient will be closer to 0 than it should be.
D) We will have a case of attenuation bias, where the coefficient is larger than it should be.

Question 12

In a case where there is multicollinearity in the model

A) Independent variables have strong linear relationships with each other
B) The variance of the estimates increases when we have multicollinearity.
C) Multicollinearity will lead to bias.
D) Both A and B
E) Both A and C

Question 13

What should we do when we have multicollinearity?

A) Ignore it no matter what.
B) Drop some of the independent variables.
C) Consider testing whether the highly collinear variables are jointly significant.
D) Add more independent variables in order to reduce multicollinearity.

Question 14

If we add more independent variables into the model:

A) The R² value will decrease if the variables we are adding into the model should not be there.
B) The R² value will increase.
C) The adjusted R² value will increase.
D) The R² will be biased.

Question 15

Including irrelevant independent variables into the model:

A) Will create bias
B) Will reduce the precision of our estimates.
C) Will have no negative or positive effect
D) Will lead to an increase in the error term.

Question 16

Which of the following is not a challenge for model specification?

A) Model fishing
B) Finding the appropriate data set
C) Changes in sample size
D) Unclear results due to the use of multiple models

Question 17

Which of the following are associated with larger omitted variable biases?

A) The dependent variable is measured with error.
B) The relationship between the included and excluded variables is weak.
C) The relationship between the included and excluded variables is strong.
D) The relationship between the excluded variable and the dependent variable is weak.

Question 18

Explain what multicollinearity is, and describe two ways to deal with the issue of multicollinearity:

Question 19

Explain goodness of fit and talk about the issue of adding irrelevant variables into the model:

Question 20

Explain the challenges in the model specification process:

Question 21

Explain the omitted variable bias problem, and show the equation(s) for determining the size of the bias.

Question 22

Explain and contrast the consequences of having a measurement error in the independent and dependent variables.