1. What is the relationship between the t and F distributions, if any?

a. The t-distribution with z degrees of freedom is F(1,z)

b. The square of the t-distribution with z degrees of freedom is F(1,z)
c. There is no relationship between the 2 distributions

2. Consider a simple linear regression function (with 1X and 1Y). Which of the following
statements is true for the estimate of sd of the coefficient of X .?
a. Proportional to the square root of ei^2
b. The ratio is positive with (Xi – X average)
c. Proportional to Xi
d. Proportional to the number of observations

a. Only (i) correct

b. Only (i) and (iv) are correct
c. Only (i), (ii) and (iv) are correct
d. Both (i), (ii), (iii) and (iv) are correct

3. Which of the following is proven by Gauss-Markov snow?

a. OLS gives an unbiased coefficient estimate
b. OLS gives the smallest estimate of the coefficient var
c. OLS only gives a minimum var estimate of the coefficient when the estimates are linear and
unbiased
d. OLS guarantees that errors (residuals) are normally distributed

4. Which of the following statements is FALSE regarding the use of p-value

in test
a. P-values are used only with two-way test
b. P-value is the exact significance level of test
c. P-value is the marginal level of significance – where we will decide between rejecting
and not rejecting the null hypothesis.
d. If we know the p-value, we can infer the statistics without looking up the statistical tables.

5. Relationship, if any, between the normal distribution and the t-distribution (student)
a. A t-distribution with 0 degrees of freedom is a normal distribution
b. A t-distribution with 1 degree of freedom is a normal distribution
c. The t-distribution with “infinite” degrees of freedom is a normal distribution
d. There is no relationship between the 2 distributions

6. Two researchers have the same model, data, coefficients and se of coefficients are
deterministic. They test the same hypothesis with a two-way test. But researcher 1 uses 5%
significance level while researcher 2 uses 10% significance level. Which of the following
assertion is true? a. Researcher 2 uses the higher observed value of t
b. Researcher 2 has a higher probability of a Type I error
c. Researcher 1 is more likely to reject the null hypothesis than d. Both researchers
will always come to the same conclusion
7. Which of the following conditions must be met for the Durbin Watson test
to be meaningful? a. The regression function must have an intercept
b. The independent regressors are not randomly selected
c. There is no lagging dependent variable in the model
d. There is no lagged independent variable in the model
1. Only (i) (ii) (iii)
2. Only (i) and (ii)
3. All
4. Only (i) (ii) and (iv)

8. What are the consequences of variable error variance?

a. Bias coefficient estimates
b. Se of the slope will be very small
c. Statistical inferences (tests and confidence intervals) may not be accurate
1. Only (i)
2. Only (ii) and (iii)
3. All
4. Only (i) and (ii)

9. In the regression model log(qi) = 2.25 – 0.7 log(pi) + 0.02yi where p is the price and q is
the quantity demanded of a certain good, y is the available income. What does the coefficient
of 0.7 mean?
a. If prices increase by 1%, the demand will be lower than the average 0.007%, while the
variable diff between materials
b. If price increases by 1%, quantity demanded will be 70% lower than average, while other
variables between the
c. If price increases by 1%, quantity demanded will be 70% less than average
d. Both the a, b, c are wrong
Used for sentences 10-12
The model is estimated using 142 observations
Y = -12.3 + 0.32 X1 + 0.043 X2 – 0.005 X3 + 0.0035 X4
Se (6.89) (0.15) (0.078) (0.0047) (0.0022)
p-val 0.07644 0.03468 0.58234 0.28928 0.11292
R^2 = 0.0395

10. What can we say about the coefficient of X4? (considered with 1-sided test)
a. If X4 increases by 1 unit, then Y increases on average 0.0035 while the other Xs stay the
same. But this is not statistically significant at 5% significance level.
b. If X4 increases by 1 unit, Y decreases on average by 0.0035 while the other Xs stay the
same. And this is statistically significant at the 5% level of significance.
c. If X4 increases by 1 unit, then Y increases on average 0.0035 and this is statistically
significant at 1% significance level.
d. The coefficient X4 is significant at the 5% level of significance but not at the 1% level of
significance.
11. The model is now estimated excluding variables X1 and X2. Then get R^2=0.0387. Value
of F-statistic for hypothesis test H0 beta1=beta2=0
a. 32.821
b. 0.0570
c. 0.0808
d. Can't calculate, need to collect more information :v

12. What can we say about the coefficients of X1 (consider the one-sided test)
a. It is statistically significant at 5% and 1% level.
b. It is statistically significant at 1% significance level.
c. It is statistically significant at 1% significance level but not at 5%.
d. It is statistically significant at the 5% level of significance but not at 1%.

13. In the reduction test for a multivariable model with Gaussian assumptions, we will increase
the probability of rejecting the hypothesis H0 that the coefficients are zero if
a. RSS of “restricted model” is larger than RSS of “unrestricted model”
b. RSS for “restricted model” is smaller than RSS for “unrestricted model”
c. Big TSS
d. The intercept is greater than the significance level
e. Both a and d are correct
f. Both c and d are correct

14. Test for normality error of residuals, obtained Chi-square(2)=31.03 with p-

value=0.0000, conclusion:
a. Normal distribution error
b. Error is not normally distributed
c. Can't conclude yet

15. Which of the following cannot be used to treat multicollinearity?

a. Remove 1 explanatory variable
b. Convert data to logarithms
c. Transform 2 explanatory variables into 1 billion ratio
d. Collect more data for all variables

16. Using Durbin watson test for the model with intercept and 3 variables X with 100
observations d=3.6. The following conclusions correct?
a. There is a positive autocorrelation
b. There is negative autocorrelation
c. No autocorrelation
d. Can't conclude yet

17. Which of the following is not a good reason to use lagged variables for
regression? a. Slow response of the dependent variable to changes in the
independent variable b. Overreaction of the dependent variable
c. The dependent variable is the central moving average of the past 4 values
d. The residuals of the model are not normally distributed
 18. What is the long-run solution to the following dynamic econometric model:
∆yt=b1 +b2∆X2t +b3∆X3t + ut
a. Y=b1 + b2X2 + b3X3
b. Yt=b1 + b2X2t + b3X3t
c. Y= -(b2/b1) X2 – (b3/b1) X3
d. There is no long-term solution to the equation

19. When adding a late dependent variable to the model, what problems will occur?
a. The assumption that the independent variables are not random is contradicted
b. Models with many lagged variables can lead to non-normally distributed
residuals c. Adding lagged variables can cause multicollinearity with existing
variables
d. se of coefficients will be wrong due to more explanatory variables

20. What are the possible consequences for the OLS estimate if there is autocorrelation?
a. Do not deviate
b. Unstable
c. Inefficient
d. Both the a, b, c are correct

21. What are the consequences of variable variance? (heteroscedasticity)

a. Estimates of the coefficients can be wrong (Ước lượng của các hệ số có thể sai)
b. Tests may give false conclusions
c. Predictions from the model may be biased
d. (Se) may not be suitable
a. (ii) and (iv)
b. (i) and (iii)
c. (i) (ii) (iii)
d. All
Khi có hiện tượng PSSS thay đổi, nếu vẫn dùng OLS để ước lượng mô hình, các ước lượng OLS thu
được vẫn là ước lượng tuyến tính, không chệch nhưng có phương sai bị chệch.

22. If negative autocorrelation occurs, which of the following is consistent with Durbin
Watson’s d? (a) close to 0
(b) close to 1
(c) close to 2
(d) close to 4

23. If autocorrelation occurs for a model with a lagged dependent variable, what might
happen?
(a) Estimating the coefficient of bias and constant (consistent)
(b) Estimating the coefficient of bias and change (inconsistent)
(c) Unbiased but variable coefficient estimate
(d) Unbiased, constant, and inefficient coefficient estimates

24. Which is not a sign of multicollinearity?

(a) large R^2
(b) When a certain variable is deleted, the regression results will change greatly
(c) The Confidence interval is narrow for coefficient estimates → large
(d) Coefficient estimates are not significant

25. If the model has inappropriate variables, the coefficient estimates will
(a) constant, unbiased but inefficient (không đổi, không chệch nhưng không hiệu quả)
(b) unchanged but deviated
(c) change
(d) constant, unbiased and efficient

26. Given the following regression model with 64 observations:

Yt=b1 + b2X2t + b3X3t + b4X4t + ut
Which of the following hypothesis H0 can be used F-test>
(i) b2=0
(ii) b2=1 And b3+b4=1
(iii) b3.b4=1
(iv) b2-b3-b4 = 1
(a) (i) and (ii)
(b) (ii) and (iv)
(c) I, ii, iii, iv
(d) I, ii, iv

27. Which of the following definitions best fits R^2 ?

(a) It is the ratio of the total variability of y explained by the model
(b) It is the ratio of the total variation of y to its mean that is explained by the model. (tỷ lệ của
tổng sự biến động của y so với giá trị trung bình của nó mà được giải thích bởi mô hình)
(c) It is the correlation between the predicted values of Y and the residual
(d) It is the correlation between the predicted values of Y and the mean

28. When R^2 = 1, which of the following statements is true?

(i) All data points lie on a straight line
(ii) ALL residuals are 0
(iii) All variation of y around the mean explained by the model
(iv) Horizontal Fitted line for all explanatory variables

(a) (ii) and (iv)

(b) (i) and (iii)
(c ) (I) ( ii) and iii
(d) all

29. Consider the following two models with lagged variables

Yt = b1+ b2 X2t + b3 Y(t-1) + ut
Mt = a1+ a2 X2t + a3 Y(t-1) + ut

Which of the following statements is true?

(i) RSS 2 equal model
(ii) R^2 is equal for both models
(iii) R^2 different corrections for the 2 models
(iv) F-test will be the same for both models

(a) (ii) and (iv)

(b) (i) and (iii)
© (i), (ii), (iii)
(d) All

30. Which of the following could be considered a disadvantage of using R^2 corrected?
(i) Corrected R^2 always increases as the variable is added
(ii) Corrected R^2 often leads to many significant variables and many insignificant
variables for multivariable models
(iii) adjusted R^2 cannot be compared for models with different explanatory variables
(iv) Adjusted R^2 cannot be compared for models with differently interpreted variables

(a) (ii) and (iv)

(b) (i) and (iii)
(c ) (i), (ii), (iii)
(d) all
(R^2 hiệu chỉnh tăng chậm hơn so với R^2 nên khi xảy ra đa cộng tuyến thì dùng R^2 hiệu
chỉnh)

31. Given 3 regression models with R^2 of 0.65, 0.78 and 0.91, respectively. Which model is
best? (The largest R^2 is the best)

32. With dummy variable D=0 is female, D=1 is male, X is the number of years of
experience and Y is the amount earned (income)
Yt = 856 + 108X + 549D + 26XD
Se 3.7 3.3 3.5 2.5
a) Compare the average income of the female with the average income of male
b) For each year of experience, compare the marginal value of women's income and
men's income

33. Using the White test with 3 X variables. How many explanatory variables can
the regression function have?
(a) 5
(b)11
(c) 10
(d) none

34. Y= b0 + b1 X1 + b2 X2 + b3X3 + u, cov (X1,X2)>0, b2 <0

Bias of b1 when X2 is removed?
(a) Shift down
(b) Deviate up
(c) stay the same