Session 4

Question 1

omitted variable bias occurs when: (1) omitted variable is correlated with ______ and (2) when omitted variable is a ______ of the ______ variable.

Answer 1

omitted variable bias occurs when:

(1) omitted variable is correlated with regressor
(2) when omitted variable is a determinant of the dependent variable.
e. g. if the (regressor) STR is correlated with percentage of Eng learners (omitted var) and that determines dependent var test score then OLS estimator will have OVB.

Question 2

Bias =

Answer 2

bias= B2*γ1

where this bias refers to α1-B1 (α1 being the coefficient from the shorter equation and B1 the coefficient from the longer equation)

Question 3

Sign of OVB:

Corr (x1, x2): +
B2 +
Bias =

Answer 3

Corr (x1, x2): +
B2 +
Bias +

Question 4

Sign of OVB:

Corr (x1, x2): +
B2 -
Bias = ?

Answer 4

Corr (x1, x2): +
B2 -
Bias = -

Question 5

Sign of OVB:

Corr (x1, x2): -
B2 +
Bias = ?

Answer 5

Corr (x1, x2): -
B2 +
Bias = -

Question 6

Sign of OVB:

Corr (x1, x2): -
B2 -
Bias = ?

Answer 6

Corr (x1, x2): -
B2 -
Bias = +

Question 7

If the bias leads our regression coefficient to be ____ in absolute value than it should be (i.e. if it moves us ____ from zero), we say that we are ____ the effect of X on Y.

Answer 7

If the bias leads our regression coefficient to be larger in absolute value than it should be (i.e. if it moves us away from zero), we say that we are overstating (or overestimating) the effect of X on Y

Question 8

Knowing that the bias has a negative sign, this means that α1 ___ B1, where B1 is the true estimate.

Knowing that the bias has a positive sign, this means that α1 ___ B1, where B1 is the true estimate

Answer 8

Knowing that the bias has a negative sign, this means that α1 B1, where B1 is the true estimate

Question 9

A higher R2 means better prediction of:.

A higher R2 does not mean greater:

Answer 9

A higher R2 means better prediction of Y using the Xs.

A higher R2 does not mean greater internal validity (i.e., is not related to bias).

Question 10

2 solutions to handle variables missing at random (MAR)

Answer 10

(1) Define an indicator variable (1 or 0) to represent when a variable is missing for a particular observation. This helps identify if there is a situation of missing at random
(2) imputation

Question 11

In the sample regression, the coefficient on ___ measures the change in __ given a one-unit increase in ___ holding all other explanatory variables constant

Answer 11

In the sample regression, the coefficient on X1i measures the change in Yi given a one-unit increase in X1i holding all other explanatory variables constant

Question 12

How do we interpret a low R2?

A higher R2 does/does not mean better prediction of Y using the Xs.
A higher R2 does/does not mean greater internal validity (i.e., is not related to bias).

Answer 12

A higher R2 does mean better prediction of Y using the Xs.

A higher R2 does not mean greater internal validity (i.e., is not related to bias).

Question 13

In order to avoid the dummy variable trap, you should….

Answer 13

In order to avoid the dummy variable trap, you should clearly define what your base case is (situation in which indicators are equal to zero).

Question 14

The Poisson distribution is a commonly observed ______ probability distribution, often observed for ______

Example:

Answer 14

The Poisson distribution is a commonly observed discrete probability distribution, often observed for count data

Examples: mortality of infants in a city, the number of misprints in a book, the number of bacteria on a plate, the number of activations of a Geiger counter, and the number of mistakes the professor makes each lecture