You are on page 1of 16

1.

A laboratory blood test is 95% eective in detecting a certain disease when it is, in
fact, present. However, the test also yields a false positive result for 1% of the
healthy people tested. If 0.5% of the population actually has the disease, what is the
probability a person has the disease given that the test was positive? Dene the events:
D: the tested person has the disease
E: the test result is positive
What do we want to compute? P(D|E)
From the problem statement, what do we know about the probabilities?
P(E|D) = 0.95, P(E|D
c
) = 0.01 P(D) = 0.005 P(D
c
) = 0.995
Using the above, we can solve for the desired probability by applying Bayes Rule:
P(D|E) =
P(E|D)P(D)
P(E|D)P(D) +P(E|D
c
)P(D
c
)
=
(0.95)(0.005)
(0.95)(0.005) + (0.01)(0.995)
0.323
So only 32% of those people whose test results are positive actually have the disease!
1
2. Suppose that of the potential new patients that call up a medical clinic, only 60%
are able to immediately speak to someone. The other 40% leave a message with their
phone numbers, in which case 75% of the time their call is returned the same day,
while 25% of the time it is returned the following day.
Data indicates that the probability that a caller will visit the clinic is 80% if their call
was picked up immediately, whereas it is 60% if the call was returned the same day,
and 40% if it was returned the following day.
For the following problems, simplify it as much as possible but your nal answer does
not have to be a single number
(a) (20pts) What is the probability that a random caller will visit the clinic?
Dene the events:
V : the event that a caller visits the clinic
I: the event that a caller spoke to someone immediately
S: the event that a caller had their call returned the same day
F: the event that a caller had their call returned the following day
From the problem statement, we have that for those who spoke to someone im-
mediately:
P(I) = 0.60 P(V |I) = 0.80
and for those who did not speak to someone immediately:
P(S|I
c
) = 0.75 P(F|I
c
) = 0.25
and
P(V |S I
c
) = 0.60 P(V |F I
c
) = 0.40
Using the denition of conditional probability we have that:
P(S I
c
) = P(S|I
c
)P(I
c
) = (0.75)(0.40)
P(F I
c
) = P(F|I
c
)P(I
c
) = (0.25)(0.40)
Using the law of total probability we therefore have that:
P(V ) = P(V |I)P(I) +P(V |S I
c
)P(S I
c
) +P(V |F I
c
)P(F I
c
)
= (0.80)(0.60) + (0.60)(0.75)(0.40) + (0.40)(0.25)(0.40)
= 0.70
2
(b) (5pts) Given that a caller visited the clinic, what is the probability that they
spoke to someone immediately when they called the clinic?
P(I|V ) =
P(V |I)P(I)
P(V )
=
(0.80)(0.60)
(0.80)(0.60) + (0.60)(0.75)(0.40) + (0.40)(0.25)(0.40)
= 0.686
3
3. An airline nds that 4 percent of the passengers that make reservations on a particular
ight will not show up. Consequently, their policy is to sell 100 reserved seats on a
plane that has only 98 seats. Find the probability that every person who shows up for
the ight will nd a seat available.
Let X count the number of passengers that show up. Then X is binomial (n =
100, p = 0.96) random variable. The probability that everybody who shows up has a
seat is:
P(X 98) = 1 P(X > 98)
= 1 P(X = 99) P(X = 100)
= 1
_
100
99
_
(0.96)
99
(0.04)
1

_
100
100
_
(0.96)
100
(0.04)
0
= 0.913
4
4. A buyer must decide whether or not to accept a large lot of items which, unknown
to him, contains 5% defectives. The lot will be accepted if there are only 0 or 1
defective items in a sample of 20 items randomly drawn from the lot. Calculate the
exact probability that the lot will be accepted.
Let X count the number of defectives in the 20 items sampled. Dene p = 0.05, then
X is a binomial(n = 20, p = 0.05) random variable, and the probability that the lot is
accepted is just:
P(X = 0) +P(X = 1) =
_
20
0
_
p
0
(1 p)
20
+
_
20
1
_
p
1
(1 p)
19
= P(X 1) = BINOMDIST(1, 20, 0.05, 1) = 0.736
5
5. Experience shows that 20% of the people reserving tables at a certain restaurant never
show up. If the restaurant has 30 tables and takes 32 reservations, use the Bernoulli
trials model to:
(a) determine the probability that the restaurant will be able to accomodate everyone.
We think of each person showing up or not as an independent Bernoulli trial
with probability of success p = 0.80. Since there are 32 reservations, we take
the number of trials to be n = 32. We can therefore model the number of people
showing up as a binomial random variable, X, with parameters (n = 32, p = 0.80).
The probability that the restaurant can accomodate everyone is thus:
P(X 30) = 1 P(X = 31) P(X = 32)
= 1
_
32
31
_
(0.80)
31
(0.20)
1

_
32
32
_
(0.80)
32
(0.20)
0
(b) Find the expected number of reservations that will be fullled.
This is the mean of a binomial r.v. with parameters (n = 32, p = 0.80):
E[X] = np = 32 0.80 = 25.6
6
6. Suppose that the number of accidents occurring on a highway each day is a Poisson
r.v with parameter = 3
(a) Find the probability that 3 or more accidents occur today (I want a number)
Let X count the number of accidents that occur. Then X is a Poisson r.v. with
parameter = 3.
P(X 3) = 1 P(X < 3)
= 1 P(X = 0) P(X = 1) P(X = 2)
= 1 e
3
3
0
0!
e
3
3
1
1!
e
3
3
2
2!
= 0.576
(b) Repeat part(a) under the assumption that at least 1 accident occurs today (again,
your answer should be a number)
P(X 3|X 1) =
P(X 3, X 1)
X 1
=
P(X 3)
P(X 1)
and:
P(X 1) = 1 P(X = 0) = 1 e
3
= 0.95
Thus:
P(X 3|X 1) =
0.576
0.95
= 0.607
7
7. The time (in minutes) until the next email arrives in your inbox is modeled as an
exponential random variable. It is observed that the probability that it takes more
than 2 minutes for the next email to arrive is exactly
3
5
the probability that it more
than 1 minute for it to arrive.
Given this information, nd the probability that it takes more than 4 minutes for the
next email to arrive.

Let X be the time until the next email arrives. We know that X is an exponential
random variable, but we do not know the parameter, , of the distribution. However,
we can solve for it from the relationship:
3
5
P(X > 1) = P(X > 2)
3
5
(1 P(X 1)) = 1 P(X 2)
3
5
e

= e
2
3
5
= e

take ln of both sides


ln
_
3
5
_
=
which yields that = ln
_
3
5
_
. Now, we can solve the problem:
P(X > 4) = 1 P(X 4) = e
4
= e
4 ln(
3
5
)
=
_
3
5
_
4
8
8. (20pts) Suppose that the weight of a person selected at random from some population
is normally distributed with parameters and . Suppose also that P(X 160) = 0.5
and P(X 140) = 0.25.
(a) (10pts) Find and

P(X 160) =
_
160

_
= 0.5
where () is the standard normal cdf. From Excel, we have that (0) = 0.5.
Therefore:
160

= 0
or = 160. We also have that:
P(X 140) =
_
140

_
=
_
140 160

_
= 0.25
Using Excel, we see that NORMSINV(0.25)=-0.67449. Therefore:
140 160

= 0.67449
We therefore obtain that = (140 160)/0.67449 = 29.65.
(b) (5pts) Find P(X 200).

P(X 200) = 1P(X < 200) = 1


_
200 160
29.65
_
= 1(1.35) = 10.9115 = 0.0885
9
(c) (5pts) Of all the people in the population weighing at least 200 pounds, what
percentage will weigh over 220 pounds?

P(X 220|X 200) =


P(X 220, X 200)
P(X 200)
=
P(X 220)
P(X 200)
=
1
_
220160
29.65
_
0.0885
=
1 (2.0236)
0.0885

1 0.9785
0.0885
= 0.243
10
9. (20pts) Suppose that the width of an item produced by a factory is normally distributed
with = 0.9000 and = 0.0030. The item is considered defective unless its width is
in the specication limits 0.9000 0.0050.
(a) (10pts) What percentage of items produced will be defective?

Let X be the width of a randomly selected item. Then X N(,


2
) with
= 0.9000 and = 0.0030.
An item is defective if X < 0.90000.0050 = 0.0050 or X > 0.9000+0.0050 =
0.0050. The probability that a random item is defective is thus:
P(X < 0.0050) +P(X > + 0.0050)
= P(X < 0.0050) + 1 P(X + 0.0050)
=
_
0.0050

_
+ 1
_
+ 0.0050

_
=
_

0.0050

_
+ 1
_
0.0050

_
= 2
_
1
_
0.0050

__
since (x) = 1 (x)
= 2
_
1
_
5
3
__
= 2(1 0.95221) = 0.095581
where the excel function NORMSDIST(
5
3
) was used to evaluate the standard
normal cdf.
The percentage of defective items is just the probability that a randomly selected
item is defective, so the answer is 9.5581 %.
(b) What is the maximum allowable value of that will permit no more than 1 in
100 defectives when the widths are normally distributed with = 0.9000 and ?

We want to determine the maximum value of such that this inequaility holds:
1
100
= 0.01 2
_
1
_
0.0050

__
The larger , is the more likely the item will be defective ( is a measure of
variability, so increasing it will increase the probability of the width being larger
than the specication limits), so the maximum value of is the one that satises:
0.01 =2
_
1
_
0.0050

__
Rewrite this as:

_
0.0050

_
= 1 (0.01/2) = 0.995
11
then use the Excel function NORMSINV(0.995) which yields 2.5758. Set this
equal to 0.005/ and solve for to obtain that:
= 0.005/2.5758 = 0.001941
This is the maximum allowable standard deviation for the width of the items
produced if we want no more than 1% of them to be defective.
12
10. (20pts) The gross weekly sales at a certain restaurant is a normal random variable
with mean $2200 and standard deviation $230. Assume the gross sales in each week
are independent of the sales in any other week.
For this problem, it is not necessary to provide a numerical answer. However, to get
full credit, be sure to dene any symbols, variables, or random variables you introduce
such that I could get a numerical answer with a calculator or computer within a few
seconds
(a) (10pts) What is the probability that the total gross sales over the next 2 weeks
exceeds $5000?
The sum of 2 independent normal random variables is itself a normal random
variable with mean
1
+
2
and variance
2
1
+
2
2
. In this case
1
=
2
= $2200
and
1
=
2
= $230. Let X N(4400, 2(230
2
)) be the total gross sales for the
next 2 weeks. The probability that total gross sales for the next 2 weeks exceeds
$5000 is then given by:
P(X > 5000) = 1 P(X 5000)
= 1
_
5000 4400

2(230)
_
= 1 (1.844)
= 0.0325
(b) (10pts) What is the probability that weekly sales exceed $2000 in at least 2 of the
next 3 weeks?
Let Y N(2200, 230
2
) be the sales for a given week. Dene p by:
p = P(Y > 2000) = 1 P(Y 2000) = 1
_
2000 2200
230
_
= 0.808
Now let W count the number of times sales exceeds $2000 in the next 3 weeks:
W is a binomial(n = 3, p) random variable. We want the probability that that
weekly sales exceed $2000 in at least 2 of the next 3 weeks. This is equivalent to:
P(W 2) = P(W = 2) +P(W = 3)
=
_
3
2
_
p
2
(1 p)
1
+
_
3
3
_
p
3
(1 p)
0
= 0.9033
13
11. (20pts) One has 100 light bulbs whose lifetimes are independent exponential random
variables with mean 5 hours. If the bulbs are used one at a time, with a failed bulb
being immediately replaced by a new one, what is the probability that there is still a
working bulb after 525 hours?
Note that if X
i
is an exponential random variable with mean 5 hours, we know that
= 1/5 = 0.20. We want to determine
P(X
1
+X
2
+ X
100
525)
that is, the probability that the sum of the lifetimes of all 100 bulbs is at least 525
hours. In terms of the sample mean S
100
=

100
i=1
X
i
/100, this is just:
P(100 S
100
525) = P(S
100
5.25)
Since the light bulbs are used independently, and assuming that n = 100 is large
enough, we can invoke the central limit theorem and approximate the probability with
the standard normal cdf:
P
_
S
n
E[S
n
]
_
Var(S
n
)
a
_
(a)
In our case, E[S
n
] = 5 and Var(S
n
) = Var(X
i
)/n = 5
2
/n = 25/n, where n = 100. We
then have that:
P(S
100
5.25) =
_
5.25 5
_
25/100
_
= (0.50)
Therefore:
P(X
1
+X
2
+ X
100
525) = 1 (0.5) = 0.3085
14
12. From past experience a professor knows that the test score of a student taking his nal
examination is a random variable with mean 75 and variance 25. Use the central limit
theorem to determine how many students would have to take the examination so as to
ensure, with probability at least 0.90, that the class average would be within 5 of 75.
Let S
n
=

n
i=1
X
i
/n. In our case E[S
n
] = 75 and Var(S
n
) = 25/n.
We use the Central Limit Approximation, such that:
P
_
S
n
E[S
n
]
_
Var(S
n
)
a
_
(a)
where S
n
N(75, 25/n). The average of the class is dened to be S
n
= (X
1
+X
2
+
+X
n
)/n, and we want to nd n such that:
P(70 < S
n
80) 0.90
We can rewrite this as:
P(S
n
80) P(S
n
70) 0.90
In terms of the standard normal cdf, this is just:
P(S
n
80) P(S
n
70) =
_
80 75
_
25/n
_

_
70 75
_
25/n
_
= (

n) (

n) = 2(

n) 1
We want to nd n such that this is at least 0.90. To nd the minimum n, set the
probability to 0.90:
2(

n) 1 = 0.90
so
n =
_

1
(0.95)

2
= [NORMSINV(0.95)]
2
= 2.7
so the professor needs at least 3 students to take the exam.
15
13. A change in the manufacturing process of steel beams is suggested. The manufacturer
wants to know if the mean strength of the beams made by a new process is dierent
from the 42,000 psi, which is the mean strength of the beams made by the present pro-
cess. A two-sided hypothesis test on is appropriate to answer this question. Breaking
strengths are measured for a test sample of n = 32 beams made by the new process.
The sample mean for this test sample is x = 42, 196 psi and the sample standard devi-
ation is s = 500 psi. Assume that the observations are normally distributed and that
the sample size is large enough such that s . At the 10% level of signicance, test
the hypothesis that:
H
0
: = 42, 000 vs H
1
: = 42, 000
Proceed in three dierent ways:
(a) Calculate the z-statistic:
z =
x
0
/

n
=
42196 42000
500/

32
= 2.217
For = 0.10, we compare this to the critical value z
/2
= NORMSINV(0.95) =
1.645. Since |z| > z
0.05
, we reject H
0
and conclude that the mean breaking
strength of the beam made by the new process is dierent. (in fact, the test
indicates it is larger).
(b) Calculate the p-value.
p = 2(1 (|z|)) = 2(1 NORMSDIST(|z|)) = 2(1 0.987) = 0.026
Since p < , reject H
0
.
(c) Compute a 90% condence interval:
_
x z
/2

n
_
=
_
42, 196 1.645
_
500

32
__
= [42051, 42341]
Since
0
does not fall in the CI, we reject H
0
16

You might also like