1.

For fixed n, the binomial distribution is always symmetric if

A. π = 0.5
B. π > 0.5
C. π < 0.5
D. unknown
2. For fixed π, the binomial distribution is more symmetric as n
A. fixed
B. increases
C. decreases
D. unknown
3. One of the following is not the assumption of a binomial distribution?
A. Fixed trials
B. Dependent trials
C. Fixed probability of success
D. Two mutually exclusive outcomes
4. For which distribution, the mean and variance are equal?
A. Binomial distribution
B. Multinomial distribution
C. Poisson distribution
D. Bernoulli distribution
5. Which one has more than two mutually exclusive outcomes?
A. Binomial distribution
B. Multinomial distribution
C. Poisson distribution
D. Bernoulli distribution
6. Which one is true about the Poisson distribution?
A. It approaches normality as µ increases
B. It approximates binomial if n is small and π is large
C. It approaches normality as µ decreases
A. unknown
7. Which test uses a known standard error (the null standard error)?
A. Score Test
B. Wald Test
 C. Likelihood-Ratio Test
D. Chi-Square Test
8. Which test is the least reliable when the sample size is very small?
A. Score Test
B. Wald Test
C. Likelihood-Ratio Test
D. Chi-Square Test
9. Which experiment has only one trial?
A. Binomial experiment
B. Multinomial experiment
C. Poisson experiment
D. Bernoulli experiment

10. Which test does not use a binomial proportion?

A. Score Test
B. Wald Test
C. Likelihood-Ratio Test
D. Poisson Test
11. Which test uses both the hypothesized standard error and the estimated standard error?
A. Score Test
B. Wald Test
C. Likelihood-Ratio Test
D. Poisson Test
12. Let X and Y denote two categorical variables having I and J categories, respectively. The
marginal distribution of each category of X variable is the sum of
A. The joint probabilities over all categories of Y variable.
B. The joint probabilities over all categories of X variable.
C. The conditional probabilities over all categories of Y variable.
D. The conditional probabilities over all categories of X variable
13. Suppose there are 150 patients in a certain hospital. Of these 60 % are females. 40% of these
patients were defaulted including 20 males. If a patient is selected at random, then what dose
the probability represents if
i. The selected patient is a female and defaulter
 A. conditional probabilities
B. joint probabilities
C. marginal distribution
D. independent probabilities
ii. The selected patient is a female?
A. joint probabilities
B. conditional probabilities
C. marginal distribution
D. independent probabilities
iii. The selected patient is defaulter if the member is female
A. joint probabilities
B. independent probabilities
C. marginal distribution
D. conditional probabilities
14. Two categorical response variables are defined to be independent if all joint probabilities are
the product of their
A. Conditional probabilities.
B. joint probabilities
C. independent probabilities
D. Marginal probabilities.
15. If two categorical response variables X and Y are independent then each conditional
distribution of Y is identical to the
A. Marginal distribution of X.
B. Marginal distribution of XY.
C. Marginal distribution of Y.
D. Joint distribution of XY
16. The row and column totals of the joint probabilities are called
A. Conditional distributions
B. Marginal distributions
C. Independent distributions
D. Common distributions
17. A relative risk of 4 has a strong association of two variables than
A. relative risk of 0.50
B. relative risk of 0.25.
C. relative risk of 6.00.
D. relative risk of 0.10
18. Which one is approximately normally distributed for large values of sample size, n.
A. logarithm o f Relativ Risk , {log( ^
RR)]
B. Relativ Risk ( ^
RR ¿,
C. ^
SE[log ( ^
RR )]
SE[log ( ^
D. CI of ^ RR )]
19. If, the odds ratio of success for row 1 to row 2 is 0.5, then the odds of success
A. in row 2 is 0.5 times the odds of success in row 1
B. in row 2 is 2 times the odds of success in row 1
C. in row 1 is 2 times the odds of success in row 2
D. in row 1 is equal the odds of success in row 2
20. The sample odds ratio θ^ is used to estimate population odds ratio θ which is given by
n11 n12
A. n n
22 21

n
¿
B. n ¿
2+¿ n ¿21

n11 n22
C. n n
12 21

n
¿
D. n ¿
1+¿ n ¿21

21. If the relative risk is above 1, then the probability that the outcome present is
A. lower for group 1 than group 2
B. equal for group 2 as group 1
C. higher for group 2 than group 1
D. higher for group 1 than group 2
22. One of the following is used to test for a small sample size in a 2x2 contingency table?
A. Pearson Chi Square Test
 B. Likelihood-Ratio Test
C. Fisher Exact Test
D. Wald Test
23. One of the following is not used to test for independence of ordinal data?
A. Linear Gamma measure
B. Pearson Chi Square Test
C. Linear Trend Measure
D. None
24. One of the following is used to test the strength of association between two categorical
variables?
A. Relative Risk
B. Wald Test
C. Likelihood-Ratio Test
D. Score Test
25. One of the following is used to test the direction of association between two categorical
variables?
A. Linear Trend Test
B. Fisher Exact Test
C. Likelihood-Ratio Test
D. Pearson Chi Square Test
26. A success is more likely to occur than a failure if

A. OR > 1
B. OR < 1
C. Log(OR)< 0
D. Log(OR) > 1

27. The estimated logit model of the probability of success is given by

A. ¿
B. α +β x i
C. exp ⁡( α + β xi )
 D. exp ⁡(β)

28. Odds Ratio (OR) of the logit model is equal to

A. exp ( α + β x i )
B. α +β x i
C. exp ⁡(β)
D. ¿

29. In a simple logistic regression model, one is not possible to be considered?

A. One predictor with two categories.
B. One response with two categories
C. One continuous predictor
D. Two variables with two categories
30. If a predictor X has six categories, then how many dummy variables are needed?
A. Seven
B. Six
C. Five
D. Four
31. The logit model of the probability of success is given by logit ^π ( x i ¿ = 3.5+1.5 age .
A. The odds of having the outcome is 4.48 times larger for one year older an individual
B. The odds of having the outcome is 33.12 times larger for one year older an individual
C. The odds of having the outcome is 4.48 times lower for one year older an individual
D. The odds of having the outcome is 1.5 times larger for one year older an individual
32. Let a model, logit ^π ( x i ¿ = -2.5+0.2 age . The probability of success at age of 20 year is
A. 0.82
B. 4.48

C. 1.5

D. 0.98
33. Let a model, logit ^π ( x i ¿ = -2.5+0.2 age . Odds of probability of success at age of 20 year is
A. 0.817
 B. 4.481
C. 1.5
D. 0.08
34. Consider the logit model of probability of success , logit ^π ( x i ¿ = 0.7−0.5 male
A. The odds of having the outcome is 0.7 times larger for females than males
B. The odds of having the outcome for males is 0.61 times that of females
C. The odds of having the outcome for males is 1.65 times that of females
D. The odds of having the outcome for females is 0.5 times that of males
35. Consider the simple logistic model, logit[ π^ ( x i ) ] = α^ + ^β x i. For a fixed, x i = x 0, a (1-α) 100%
confidence interval for logit[ π ( x 0 ) ] is

A. exp (α^ + ^β x 0 )± z Ə/ 2 √ Var ( α^ + ^β x 0 ¿)¿.

B. exp (α^ + ^β x 0 )± z Ə/ 2 √ Var ¿ ¿

C. (α^ + ^β x 0 )± z Ə √ Var ¿ ¿
2

D. (α^ + ^β x 0 )± z Ə/ 2 √ Var ( α^ + ^β x 0 ¿)¿.

36. A target variable has three or more possible values without any order in
A. Binary Logistic Regression
B. Ordinal Logistic Regression
C. Multinomial Logistic Regression
D. Multiple Logistic Regression
37. How many different types of logistic regressions are there?
A. 2
B. 3
C. 4
D. 5
38. Which of the following is used where the target variable is of categorical nature?
A. Poisson Regression
B. Linear Regression
C. Logistic Regression
D. Binary Regression
39. The target variable with pass or fail values is an example of?
A. Binary Logistic Regression
B. Multinomial Logistic Regression
C. Multiple Logistic Regression
D. Ordinal Logistic Regression
40. Suppose, a British study reported that of smokers who get lung cancer, “women were 1.7
times more vulnerable than men to get lung cancer. Then 1.7 represents
A. Odds ratio
B. Prevalence rate
C. Incidence rate
D. Relative risk
41. In the two contingency table, the relative risk and odds ratio take similar values if the
proportion in the first category for each group is close to
A. One
B. Half
C. Double
D. Zero
42. If the probability of successes are equal in the two groups being compared and no association
between the variables, then
A. Relative Risk = 0
B. log (Relative Risk) = 0
C. Relative Risk = 0.5
D. log (Relative Risk) = 1
43. It is plausible that there is an effect of explanatory variable on the response variable if
A. Confidence interval for log odds ratio is (-0.5, 1.3)
B. Confidence interval for log odds ratio is (0.3, 1.8)
C. Confidence interval for odds ratio is (0.5, 1.4)
D. Confidence interval for odds ratio is (-0.3, 1.2)
44. We found that a 95% confidence interval for the odds ratio relating having a heart attack
(yes, no) to drug (placebo, aspirin) is (1.44, 2.33). If we had formed the table with aspirin in
the first row (instead of placebo), then the 95% confidence interval would have been
 A. (2.33,1.44)
B. (1.44, 2.33)
C. (1/2.33, 1/1.44)
D. (1/1.44, 2. 33)
45. If the categories of the outcome variable have a natural order, then the appropriate model is
A. Multinomial logistic regression
B. Ordinal logistic regression
C. Binary logistic regression
D. Simple logistic regression
46. If the outcome variable Y has 4 response categories without natural order then how many
models should be performed?
A. One
B. Two
C. Three
D. Four
47. The overall significance of predictors in the logistic regression model is tested by
A. Likelihood Ratio Test
B. Wald Test
C. Linear trend test
D. Wald Test
48. If the outcome variable Y has 3 response categories with natural order then we must perform
A. One cumulative logit and one intercept
B. Two cumulative logit and three intercepts
C. Three cumulative logit and two intercepts
D. Two cumulative logit and two intercepts
49. For ordinal logit model, the parallel regression assumption implies that
A. The effect of each predictor is the same for all cumulative logit
B. The effect of each predictor is different for each cumulative logit
C. The effect of each predictor is different for some cumulative logit
D. The effect of each predictor is the same for some cumulative logit
50. Let a National Cancer Institute study about breast cancer reported that the women taking the
drug were 45% less likely to experience breast cancer than women taking the placebo.
A. The relative risk for those taking drug compared to those taking placebo is 0.45
B. The relative risk for those taking placebo compared to those taking drug is 0.55
C. The relative risk for those taking drug compared to those taking placebo is 0.55
D. The relative risk for those taking drug compared to those taking placebo is 1.55
51. Which mode is applied to analysis the count response data?
A. Binary Logistic Regression
B. Ordinary Logistic Regression
C. Multinomial Logistic Regression
D. Poisson Regression Model
52. A model fit predicting preference for president (Democrat, Republican, Independent) using
π^ D π^ R
X = daily income (in birr) is Log ( ) = 3.3-0.2X and Log ( ) = 1.0 + 0.3X.
π^ I π^ I
π^ R
I. The prediction equation for Log ( ) is
π^ D
A. -2.3+0.1X
B. -2.3+0.5X
C. 2.3-0.5X
D. 2.3-0.1X
II. The value of income at which two estimated probabilities are the same¿ π^ D) is
A. 4.6 birr
B. 3.2 birr
C. 6.4 birr
D. 2.3birr
III. if the daily income is 2 birr, then the estimated probabilities of Independent( ^π I ) is
A. 0.035
B. 0.052
C. 0.168
D. 0.041
53. If the two tailed test statistics, Z cal value is 2, then the corresponding P- value is (use:
P(0<Z<2) = 0.4772)
A. 0.0228
B. 0.0456
C. 0.5228
D. 0.4772
54. The table below shows the output report for a Poisson loglinear model using X = weight and
Y= number of satellites
Parameter estimate ( ^β ) std error 95% conf Limits Chi- Sq Pr > Chisq
Intercept -0.4284 0.1789 -0.7791 -0.0777 5.753 0.0167
Weight 0.5893 0.0650 0.4619 0.7167 82.15 <.0001
I. The fitted loglinear model of the response is
A. Exp(-0.4284+0.5893X)
B. -0.4284+0.5893X
C. Exp(0.5893X)
D. 0.5893X
II. The estimated average number of satellites at the weight of 2.5 kg is
A. 2.8429
B. 1.0448
C. 4.3634
D. 1.4732
III. The confidence interval for the true weight effect(β) is
A. (1.5871,2.0477)
B. (1.6893,1.9238)
C. (0.4619, 0.7167)
D. (0.2539,12.7981)
IV. One of the following is true between the estimated number of satellites and weight?
A. The estimated number of satellites is independent on weight
B. There is indirect relation between the average number of satellites and weight
C. The estimated average number of satellites increases by a factor of epx(0.0650)
D. The estimated average number of satellites is dependent on weight.
55. Does marital happiness depend on family income? For the 2002 General Social Survey,
counts in the happiness categories as (not happy, pretty happy, very happy) for below
average income, for average income and for above average income (1,2,3) respectively.
Table below shows output for a cumulative logit model with very happy as the baseline
category with scores {1, 2, 3} for the income categories.
Parameter Estimate ( ^β ) Std error Chi- Sq Pr > Chisq DF
Intercept 1 -3.2467 0.3404 90.9640 < 0.001 1
Intercept 2 -0.2378 0.2592 0.8414 0.3590 1
Income -0.1117 0.0018 6.9316 < 0. 05 1
I. Based on the information in the table, one of the following is false?
A. There are two cumulative probabilities
B. There are two intercept parameters
C. There are two slop parameters
D. There is one slop parameter
II. Based on the information in the table, one of the following is true?
A. The estimated odds of being less happy increases as income increase.
B. The estimated odds of being less happy decreases as income increase
C. The estimated odds of being pretty happy increases as income increase.
D. The effect of income is different for the odds of being less happy and pretty happy
III. The probability that a person with average family income reports not happy marriage(
^π 1) is
A. 0.39
B. 0.63
C. 0.03
D. 0.36
IV. The cumulative probability that a person with average family income reports not happy
or pretty happy marriage( π^ ( Y ≤ 2 ) ) is
A. 0.39
B. 0.63
C. 0.06
D. 0.36
 V. The probability that a person with average income reports a very happy marriage( ^π 3)
A. 0.39
B. 0.36
C. 1.00
D. 0.61

56. Suppose the estimated model is given by logit [ ^π ( x i ¿ ] = 2.828+ 0.055 x i and the

variance-covariance matrix of the estimated parameters is: (8.509−0 . 179

0 . 004 )
I. The 95% confidence interval for the slop (β) is
A. (0.933; 1.196)
B. ¿)
C. (-2.567; 0.903)
D. (0.0712, 0.712)
II. The 95% confidence interval for the odds ratio(OR) is
A. (0.933; 1.196)
B. ¿)
C. (-2.567; 0.903)
D. (0.0712, 0.712)
III. The 95% confidence interval for the logit probability of success [logit( π ( x i ¿ 36.3))] is
A. ¿)
B. (0.933; 1.196)
C. (-2.567; 0.903)
D. (0.0712, 0.712)
IV. The 95% confidence interval for the probability of success π ( x i ¿ 36.3) is
A. ¿)

B. (0.933; 1.196)

C. (-2.567; 0.903)
D. (0.0712, 0.712)
57. The one with the _____________is the best model.
A. Smallest value of Bayesian Information Criteria
 B. Largest value of Akaike information criterion
C. Largest value of Bayesian Information Criteria
D. None of the above
58. In the Hosmer-Lemeshow Test, which one is true?
A. The null hypotheses states that the model is not a good fit
B. The null hypotheses states that the model is a good fit
C. The alternative hypotheses states that the model is a good fit
D. None of the above
59. Using the Hosmer-Lemeshow Test, the model is a good fitted model?
A. If the P-value of chi-square test is greater than the significance level (α)
B. If the P-value of chi-square test is smaller than the significance level (α)
C. If the P-value of chi-square test is greater than to the half of significance level (α/2)
D. If the P-value of chi-square test is smaller than to the half of significance level (α/2)
1
60. Consider the logistic function is f(h) = ; −∞ < z < ∞ . The range of f(h) is
1+ exp ⁡(−h )
A. -1 ≤ f (z) ≤ 0
B. -1≤ f (z) ≤ 1
C. 0 ≤ f (z) ≤ 1
D. 0.5 ≤ f (z) ≤ 1