MAE 108 - Probability and Statistical Methods for Engineers - Spring 2014

Final Exam, June 10

Instructions
(i) Open book, open notes, calculator with no communication capabilities are allowed,
(ii) No cellphones, tablets, or laptops,
(iii) You have 170 minutes,
(iv) Do not get stuck! Move on and come back later if you have time,
(v) Justify all your answers,
(vi) Do not forget to write your name and student number in your exam.

Questions
1. (21 points) An architecture firm takes part in three competitions. Two of them are public contests to
build the city fire department headquarters (F) and the police department headquarters (D). The third
one is a private project to design a large winnery center (W). The probability of winning any contest
is 0.25, and winning any of the public contests is statistically independent of winning the private
contest. On the other hand, if the firm wins the D, then probability of winning the F becomes 0.7.

(i) What is the probability of winning at least one public contest?

(ii) If the firm wins at least a public contest, what is the probability that it is the police department
headquarters and not the fire department headquarters?
(iii) What is the probability of winning at least one contest?
(iv) If the firm wins a contest, the probability that the firm goes bankrupt over the next year is 0.1,
while if it does not win any contest, the probability that it goes bankrupt is 0.7. What is the
probability that the firm shuts down over the next year?

Solution: Let us denote the events

F = Winning the contest to build the fire department headquarters,

D = Winning the contest to build the police department headquarters,
W = Winning the contest to design the winnery.

We know that P (F) = P (D) = P (W ) = 0.25, and that P (F|D) = 0.7.

(+ 4 points)

(i) The probability of winning at least a public contest is

P (F ∪ D) = P (F) + P (D) − P (F ∩ D) = 0.25 + 0.25 − P (F|D)P (D)

= 0.5 − 0.7 × 0.25 = 0.5 − 0.175 = 0.325 .

(+ 4 points)
 (ii) We are asked

P (D ∩ F) P (F|D)P (D) (1 − P (F|D))P (D)

P (D ∩ F|F ∪ D) = = =
P (F ∪ D) 0.325 0.325
(1 − 0.7) ∗ 0.25
= = 0.230 .
0.325

(+ 5 points)

(iii) The probability of winning a contest becomes

P (W ∪ F ∪ D) = 1 − P (W ∩ (F ∪ D)) = 1 − P (W )P (F ∪ D)
= 1 − (1 − 0.25) × (1 − 0.325) = 1 − 0.75 ∗ 0.675 = 0.49375 .

(+ 4 points)

(iv) Let us denote the event B = the firm goes bankrupt. We are told that P (B|F ∪ D ∪ W ) = 0.1
and that P (B|F ∩ D ∩ W ) = 0.7. Then,

P (B) = P (B|F ∪ D ∪ W )P (F ∪ D ∪ W ) + P (B|F ∩ D ∩ W )P (F ∩ D ∩ W )

= 0.1 ∗ 0.49375 + 0.7 ∗ (1 − 0.49375) = 0.403

(+ 4 points)
(In all the above, maintain partial credit, take off -0.25 per computational mistake)

2. (18 points) Suppose a yogurt manufacturer stamps a special coupon under the lid of a yogurt with a
probability 0.01, independently across yogurts.

(i) What is the probability that there are no more than two coupons in a set of 100 yogurts?
(ii) How many yogurts do you have to buy to make sure that with a 95% probability you will find
at least one coupon among the set?
(iii) What probability should the manufacturer employ to print a coupon so that the chance that a
customer finds at least one coupon in a set of 100 is reduced to 0.01?

Solution:

(i) We can model the number of coupons X in the 100 yogurt set as a Binomial distribution
with parameters n = 100 and p = 0.01.
(+ 2 points)

Page 2
 The probability that there are no more than two coupons in a set of 100 is

P (X ≤ 2) = P (X = 0) + P (X = 1) + P (X = 2)
     
100 0 100 100 1 99 100
= (0.01) ∗ (0.99) + (0.01) ∗ (0.99) + (0.01)2 ∗ (0.99)98 =
0 1 2
9900
(0.99)100 + 100 ∗ (0.01)(0.99)99 + (0.01)2 (0.99)98 = 0.3660 + 0.3697 + 0.1848 = 0.9205
2

(+ 8 points)

(ii) We need to find n in the Binomial so that 0.95 = P (X ≥ 1).

(+ 2 points)
Then,
 
n
0.95 = P (X ≥ 1) = 1 − P (X = 0) = 1 − (0.01)0 (0.99)n = 1 − 1 ∗ (0.99)n
0

(+ 2 points)
From here,

0.05 = (0.99)n ⇐⇒
ln 0.05 = n ln 0.99 ⇐⇒
ln 0.05 −2.9957
n= = = 298.07
ln 0.99 −0.01005
Then, you need to buy at least 299 yogurts.
(+ 1 point)

(iii) The manufacturer needs to print coupons with probability p so that 0.01 = P (X ≥ 1).
(+ 1 point)
 
100 0
0.01 = P (X ≥ 1) = 1 − P (X = 0) = 1 − p (1 − p)100 = 1 − (1 − p)100
0

(+ 1 point)
Solving for p, we have

0.99 = (1 − p)100 ⇐⇒ ln 0.99 = 100 ln(1 − p)

( ln100
0.99
) ln 0.99
1−p=e ⇐⇒ p = 1 − e( 100
)
= 0.0001

(+ 1 points)

3. (15 points) Two highways A and B merge on a third one C. During certain hour H of the day, the
maximum traffic flow over highway A as measured by video cameras has an average of 45 cars per

Page 3
minute with a standard deviation of 10 cars per minute, whereas in highway B it has a mean of 35
cars per minute with a standard deviation of 10 cars per minute. Suppose that traffic flow in both
highways follows a normal distribution and that they are statistically independent.

(i) What is the distribution of the maximum traffic flow for highway C during hour H?
(ii) Suppose that highway C can accommodate a flow of 90 cars per minute without getting con-
gested. What is the probability that C will become congested during hour H?
(iii) If you would like to reduce the probability of congestion during hour H to 0.01, how much
would you have to expand highway C (number of cars/minute)?

Solution:

(i) The maximum traffic flow in highway A during hour H is a random variable X ∼ N (45, 10),
while the maximum traffic flow in highway B is a random variable Y ∼ N (35, 10).
(+ 2 points)
Since both highways merge in C then, its maximum traffic during hour H is Z = X + Y ∼
N (µ, σ).
(+ 2 points)
Since X, Y are independent, the parameters µ and σ are:
√
µ = 45 + 35 = 80, σ 2 = 102 + 102 = 2 × 102 =⇒ σ = 10 2

(+ 2 points)

(ii) Suppose the capacity of highway C is 90 cars per minute. Then, the probability of C becom-
ing congested is given as
 
90 − 80
P (Z > 90) = 1 − P (Z ≤ 90) = 1 − Φ √ = 1 − Φ(0.7071) ≈ 1 − 0.758 = 0.242
10 2

(+ 4 points)

(iii) To reduce congestion to 0.01, we need to expand highway C so that the number of cars z per
minute it can accommodate during the maximum traffic flow is
 
z − 80
0.01 = P (Z > z) = 1 − P (Z ≤ z) = 1 − Φ √ ⇒
10 2
 
z − 80 z − 80
0.99 = Φ √ ⇒ 2.33 = √
10 2 10 2
√
z = 80 + 23.3 2 = 112.95

(+ 5 points)

Page 4
4. (13 points) The income for an average family for students enrolled at a university is $45, 000 per year.
A donor would like to set up a scholarship fund for students in financial need as she feels that the
average family income (for families all over the state) is less than that of the university as a whole. To
test this belief, she has the administration pull 25 state family records at random. It is found that the
average income of these students is $44, 500 with a sample standard deviation of $500. Can she feel
justified that a scholarship fund needs to be set for students? (discuss results for both a 1% and a 0.1%
significance levels.)

Solution: We set up a one-sided hypothesis test, with a null and an alternative hypothesis as
follows:

H0 : µ = $45, 000
HA : µ < $45, 000 .

(+ 3 points)
Since we are testing for the mean and the population true variance is unknown, we choose the
sample mean as a test statistic with the t-distribution with n − 1 = 24 degrees of freedom.
(+ 3 points)
In this way, we have

X̄ − 45, 000
T = √ ∼ t-distribution.
500/ 25

(+ 2 points)
44,500−45,000 −500
From the data we have, we compute the value t = 100 = 100 = −5.
(+ 1 point)
With a significance level of 1%, the critical value for the lower region of rejection is t0.01,24 =
−t0.99,24 = −2.4922.
(+ 1 point)
Clearly, −5 < −2.4922 so, based on this significance level, we reject the null hypothesis and the
donor is right in setting up this scholarship.
(+ 1 point)
For a lower significance level of 0.1%, the critical value for the lower region of rejection is t0.001,24 =
−t0.999,24 = −3.4668.
(+ 1 point)
Based on −5 < −3.4668, the donor can reject the null hypothesis with higher confidence and set
up a scholarship fund.
(+ 1 point)

5. (21 points) Recorded data of temperatures over different days are 47.5, 52.3, 56.8, 48.4, 54.2, 55.1, 54.4,
50.5, 52.7, 48.8.

Page 5
 (i) Provide point estimations of the mean and variance of the temperature.
(ii) Suppose that the sample variance of the temperature is the true variance of the population. De-
termine the 92% two-sided confidence interval of the mean temperature.
(iii) Suppose the true variance of the population is unknown. Compute the 95% two-sided confi-
dence interval of the mean temperature.

Solution:

(i) Point estimations of the mean and variance are the sample mean, x̄, and the sample variance,
s2 . These are computed as follows:
1 1
x̄ = (47.5 + 52.3 + 56.8 + 48.4 + 54.2 + 55.1 + 54.4 + 50.5 + 52.7 + 48.8) = 520.7 = 52.07,
10 10
10
1 X 2 1
s2 = ( xi − 10x̄2 ) = (2256.25 + 2735.29 + 3226.24 + 2342.56+
9 9
i=1
+ 2937.64 + 3036.01 + 2959.36 + 2550.25 + 2777.29 + 2381.44 − 27112.849)
1 89.481
(27202.33 − 27112.849) = = 9.9423
9 9

(+5 points)

(ii) To compute the two-sided 92% confidence interval, we need to use α = 1 − 0.92 = 0.08, and
1 − α/2 = 1 − 0.04 = 0.96.
(+ 1 point)
√
Then, assuming that σ = s = 9.9423 = 3.1531 is the true standard deviation of the popu-
lation, the critical values to compute the confidence interval are obtained from the standard
Gaussian distribution as:

k0.96 = Φ−1 (0.96) = 1.76, k0.04 = −1.76.

(+ 3 points)
Then, the two-sided confidence interval is
 
s s
hµi0.92 = x̄ + k0.04 √ , x̄ + k0.96 √
10 10
 
3.1531 3.1531
= 52.07 − 1.76 × √ ; 52.07 + 1.76 × √
10 10
= (52.07 − 1.7549; 52.07 + 1.7549) = (50.3151; 53.8249).

(+ 4 points)

(iii) If the variance of the population is unknown, we have to resort to the t-distribution. Now
α = 1 − 0.95 = 0.05, and 1 − α/2 = 1 − 0.025 = 0.975. The critical values to compute the
confidence interval are

t0.025,9 = −t0.975,9 = −2.2622

Page 6
 (+ 4 points)
Then, the two-sided confidence interval becomes
 
s s
hµi0.95 = x̄ + t0.025,9 √ , x̄ + t0.975,9 √
10 10
 
3.1531 3.1531
= 52.07 − 2.2622 × √ ; 52.07 + 2.2622 × √
10 10
= (52.07 − 2.6144; 52.07 + 2.6144) = (49.4556; 54.6844) .

(+ 4 points)

6. (12 points) Assume the number of car accidents on a specific road follows a Poisson distribution with
a mean rate of occurrence equal of 3 accidents per day.
(i) What is the probability that there are no car accidents during a weekend on that road?
(ii) What is the probability that the total number of car accidents during a weekend will be at least 1
but not larger than 2?
(iii) Suppose that car accidents can be classified into injury (I) and non-injury (N) accidents. Assume
the number of both types of accidents are statistically independent processes that follow a Pois-
son distribution, and that the mean occurrence rate for non-injury accidents over a day is 2. What
is the mean occurrence rate of the number of injury accidents over a day?
(iv) What is the probability that over a weekend there is exactly one accident which results in injury
and one accident which results into non-injury?

Solution:

(i) A weekend has 2 days, therefore, we have that the number of car accidents over a weekend
X follows a Poisson distribution with parameter λ = v ∗ 2 = 6. The probability that there
are no car accidents during a weekend is then

λ0 −6
P (X = 0) = e = e−6 ≈ 0.0025 .
0!

(+ 3 points)

(ii) The probability is

λ1 λ2
 
P (1 ≤ X ≤ 2) = + e−6
1! 2!
62
 
= 6+ e−6 = 24e−6 = 0.0595 .
2

(+ 3 points)

Page 7
(iii) Let I be the number of car accidents which result into injury and N be the number of car ac-
cidents which result into non-injury. It holds that X = I + N , where I and N are statistically
independent.
(+ 1 point)
Since I and N are statistically independent, we have that λX = λI + λN .
(+ 1 point)
In this way we have that λI = λX − λN = 1, over a day.
(+ 1 point)

(iv) For a weekend, we need to use parameters λI,2 = 2 ∗ 1 = 2 and λN,2 = 2 ∗ 2 = 4. Since I and
N are independent, the desired probability is

(2)1 −2 (4)1 −4
P (I = 1, N = 1) = P (I = 1)P (N = 1) = e ∗ e = 8 ∗ e−6 = 0.0198
1! 1!

(+ 3 points)
(maintain partial credit, assuming λI was computed correctly)

Page 8