EDA EXAM

Question 1

Ui (error term)

Answer 1

Ui represent all factors influencing independent variable other than dependent variables

Question 2

R^2 - interpret

Answer 2

R^2 indicates variation in dependent variable predicts % of the variance of independent variable (y)
Be careful to calculate %

Question 3

R^2 - fit in the model?

Answer 3

R^2 increases as regressor is added. An increase in R^2 does not means that adding regressor actually improve fit of model

Question 4

R^2 - good model for estimating effect?

Answer 4

R^2 cannot be used to show whether there is omitted variable bias so it is not informative as to whether the model is appropriate for estimating causal effect

Question 5

Nonlinearity

Answer 5

A test for nonlinearity a test of the null hypothesis that X^2 is zero.

Question 6

Endogenous

Answer 6

1. there is Omitted variable bias
2. measurement error

Question 7

Instrumental variable

Answer 7

To be a valid instrumental variable, the variable must satisfy the instrument relevance and instrument exogeneity. The new variable requires to have nonzero coefficient in the population regression of original variable and new variable and other regressor in the mode
Exogeneity requires that it is not correlated with the error term.

Question 8

Instrumental variable - in case there is the first-staged F-statistic

Answer 8

Compare the first stage F-statistic and the rule of thumb value of 10
greater than10 - satisfy instrument

Question 9

Interpret the estimated coefficient

Answer 9

1. explain the relationship between independent and dependent variable
2. t test - the estimated coefficient is statistically significant or not
3. mention causal relationship

Question 10

Interpret the estimated coefficient - in case of one regressor

Answer 10

Given the absence of other regressor, care should be taken to avoid the use of language that implies causal relationship

Question 11

Explain difference between estimated coefficient in column 1 and 2 (ex)

Answer 11

1. examine statistically significant or not for column 1
2. examine statistically significant or not for column 2
3. This indicate column 1 is affected by omitted variable bias. Since the estimated coefficient on other variable in column 2 is positive, there must be positive correlation between new and original variable. The omission lead to an overestimate of the effect of independent and dependent variable

Question 12

NO estimate of the interpret

Answer 12

When the within estimator is used, the interpret is estimated from the model by the within transformation because it does not vary over time.

Question 13

Fixed effect estimation (FE)

Answer 13

fixed effect estimation controls of time-invariant, unobserved factors.

Question 14

Adding new variable case 1

Answer 14

New variable would be highly collinear to original variable. This would lead to imprecise of both variables. This additionally increase likelihood that t-test will lead to the conclusion parameter estimates are not significantly different from zero.

Question 15

Adding new variable case 2

Answer 15

the researchers is concerned with omitted variable bias. The omission of a variable that is correlated with an independent variable and dependent variable since it generate a correlation between and error term.

Question 16

Autocorrelation

Answer 16

A variable exhibits autocorrleation if it is correlated over time. Autocorrelation of error term is problematic in the OLS estimation while still biased, is not efficient.
If the errors are autocorrelated, then homoskedastic standard error estimates are incorrect. Consequently, inference will also be incorrect.

Question 17

Heteroskedasticity and Homoskedasticity

Answer 17

Heteroskedasticity does not affect consistency as long as other conditions of OLS consistency hold. If the error are heteroskedastic, then the homoskedasitc standard error estimates are incorrect. Consequently, inference will also be incorrect.