Math 3C Homework 9 Solutions

Ilhwan Jo and Akemi Kashiwada

ilhwanjo@math.ucla.edu, akashiwada@ucla.edu

Assignment: Section 12.6 Problems 15, 16, 17(a), 18(a), 19(a), 20(a), 23-26

15. Toss a fair coin 400 times. Use the central limit theorem to find an approximation for the probability of
at most 190 heads.
Solution
Let (
1, if ith toss is heads
Xi = .
0, otherwise
Since the Xi are i.i.d. and binomially distributed we know
µ ¶ µ ¶µ ¶
1 1 2 1 1 1
µ = EXi = np = (1) = , σ = var(Xi ) = np(1 − p) = (1) 1− = .
2 2 2 2 4
P400
Now let S400 = i=1 Xi count the number of heads we get out of 400 coin tosses. By the central limit
theorem we get
µ ¶
S400 − 400µ 190 − 200
P (S400 ≤ 190) = P √ ≤
400σ 2 10
µ ¶
S400 − 400µ
=P √ ≤ −1
400σ 2
≈ 1 − Φ(1)
= 1 − 0.8413
= 0.1587

16. Toss a fair coin 150 times. Use the central limit theorem to find an approximation for the probability that
the number of heads is at least 70.
Solution
P150
Let Xi be defined as in the previous problem and let S150 = i=1 Xi . By the central limit theorem
µ ¶
S150 − 150µ 70 − 75
P (S150 ≥ 70) = P √ ≥ √
150σ 2 37.5
µ ¶
S150 − 150µ
=1−P √ ≤ −0.82
150σ 2
≈ 1 − Φ(−0.82)
= Φ(0.82)
= 0.7939

1
17a. Toss a fair coin 200 times. Use the central limit theorem to find an approximation for the probability that
the number of heads is at least 120.
Solution
P200
Let Xi be defined above and let S200 = i=1Xi . By the central limit theorem
µ ¶
S200 − 200µ 120 − 100
P (S200 ≥ 120) = P √ ≥ √
200σ 2 50
µ ¶
S200 − 200µ
=1−P √ ≤ 2.83
200σ 2
≈ 1 − Φ(2.83)
= 1 − 0.9977
= 0.0023

18a. Toss a fair coin 300 times. Use the central limit theorem to find an approximation for the probability that
the number of heads is between 140 and 160.
Solution
P300
Let Xi be defined above and let S300 = i=1 Xi . By the central limit theorem
µ · ¸¶
S300 − 300µ 140 − 150 160 − 150
P (S300 ∈ [140, 160]) = P √ ∈ √ , √
300σ 2 75 75
µ ¶ µ ¶
S300 − 300µ S300 − 300µ
=P √ ≤ 1.15 − P √ ≤ −1.15
300σ 2 300σ 2
≈ Φ(1.15) − Φ(−1.15)
= 0.8749 − (1 − 0.8749)
= 0.7498

19a. Suppose Sn is binomially distributed with parameters n = 200 and p = 0.3. Use the central limit theorem
to find an approximation for P (99 ≤ Sn ≤ 101) without the histogram correction.
Solution
Pn
Since Sn is binomially distributed, Sn = i=1 Xi where
½
1 if the k-th trial is successful
Xk =
0 otherwise

We have

µ = EXi = 1 · p + 0 · (1 − p) = p = 0.3,
σ 2 = var(Xi ) = EXi2 − (EXi )2 = 12 · p + 02 · (1 − p) − p2 = p(1 − p) = 0.3 · 0.7 = 0.21.

By the central limit theorem,

µ ¶
99 − 200 · 0.3 Sn − nµ 101 − 200 · 0.3
P (99 ≤ Sn ≤ 101) = P √ ≤ √ ≤ √
200 · 0.21 nσ 2 200 · 0.21
µ ¶
Sn − nµ
= P 6.02 ≤ √ ≤ 6.33
nσ 2
≈ Φ(6.33) − Φ(6.02)
= 0.

20a. Suppose Sn is binomially distributed with parameters n = 150 and p = 0.4. Use the central limit theorem
to find an approximation for P (Sn = 60) without the histogram correction.
 Solution
Since Sn is binomially distributed, we know

µ = EXi = p = 0.4, σ 2 = var(Xi ) = p(1 − p) = (0.4)(0.6) = 0.24.

By the central limit theorem,

µ ¶
Sn − nµ 60 − 150 · 0.4
P (Sn = 60) = P √ = √
nσ 2 150 · 0.24
µ ¶
Sn − nµ
=P √ =0
nσ 2
≈ P (Z = 0)
=0

where Z is the standard normally distributed random variable.

23. How often should you toss a coin to be at least 90% certain that your estimate of P (heads) is within 0.1
of its true value?
Solution
Let ½
1 if ith toss is heads
Xi =
0 otherwise
Pn Pn
Then X n = n1 i=1 Xi is an estimate of the proportion of tosses resulting in heads. If we let Sn = i=1 Xi ,
√
then n √X n −p = S√nnσ
−nµ
2
is approximately standard normally distributed with µ = p = EXi , σ 2 =
p(1−p)
var(Xi ) = p(1 − p). We want to find n so that

P (|X n − p| ≤ 0.1) ≥ 0.9,

or à !
√ −0.1 √ Xn − p √ 0.1
P np ≤ np ≤ np ≥ 0.9.
p(1 − p) p(1 − p) p(1 − p)
√
Since n √X n −p is approximately standard normally distributed,
p(1−p)

à !
√ 0.1
2Φ np − 1 ≥ 0.9
p(1 − p)
or à !
√ 0.1
Φ np ≥ 0.95
p(1 − p)
which is true if
√ 0.1
np ≥ 1.65
p(1 − p)
So we have
n ≥ (16.5)2 p(1 − p)
1
where p(1 − p) attains its maximum 4 when p = 12 . This gives us n ≥ 68.06, which implies n = 69 would
be sufficient.
24. How often should you toss a fair coin to be at least 90% certain that your estimate of P (heads) is within
0.01 of its true value?
Solution This question is basically same as 23. So we have
à !
√ 0.01
Φ np ≥ 0.95
p(1 − p)

which is true if
√ 0.01
np ≥ 1.65
p(1 − p)
So we have
n ≥ (165)2 p(1 − p)
1 1
where p(1 − p) attains its maximum 4 when p = 2. This gives us n ≥ 6806.3, which implies n = 6807
would be sufficient.
25. To forecast the outcome of a presidential election in which two candidates run for office, a telephone poll
is conducted. How many people should be surveyed to be at least 95% sure that the estimate is within
0.05 of the true value?
Solution
We assume that each individual votes for one and only one of the two candidates. Let A be one of the
two candidates and ½
1 if ith individual voted for A
Xi =
0 otherwise
Pn
We assume that Xi ’s are i.i.d. Then X n = n1 i=1 Xi is an estimate of the proportion of individuals
Pn √
who voted for A. If we let Sn = i=1 Xi , then n √X n −p = S√n −nµ
2
is approximately standard normally
nσ
p(1−p)
distributed with µ = p = EXi , σ 2 = var(Xi ) = p(1 − p). We want to find n so that

P (|X n − p| ≤ 0.05) ≥ 0.95,

As in previous questions, we get

à !
√ 0.05
2Φ np − 1 ≥ 0.95
p(1 − p)
or à !
√ 0.05
Φ np ≥ 0.975
p(1 − p)
which is true if
√ 0.05
np ≥ 1.96
p(1 − p)
So we have
1
n ≥ (39.2)2 · = 384.2,
4
which implies n = 385 would be sufficient.
26. A medical study is conducted to estimate the proportion of people suffering from seasonal affected disorder.
How many people should be surveyed to be at least 99% sure that the estimate is within 0.02 of the true
value?
Solution
Let ½
1 if ith individual is suffering from the disorder
Xi =
0 otherwise
 1
Pn
We assume that Xi ’s are i.i.d. Then X n = nXi is an estimate of the proportion of individuals
i=1
Pn √
who is suffering from the disorder. If we let Sn = i=1 Xi , then n √X n −p = S√nnσ
−nµ
2
is approximately
p(1−p)
standard normally distributed with µ = p = EXi , σ 2 = var(Xi ) = p(1 − p). We want to find n so that

P (|X n − p| ≤ 0.02) ≥ 0.99,

Again, we get à !
√ 0.02
2Φ np − 1 ≥ 0.99
p(1 − p)
or à !
√ 0.02
Φ np ≥ 0.995
p(1 − p)
which is true if
√ 0.02
np ≥ 2.58
p(1 − p)
So we have
1
n ≥ (129)2 · = 4160.3,
4
which implies n = 4161 would be sufficient.