1

Course of Study Exercises Statistics

Bachelor Computer Science WS 2020/21

Sheet VIII - Solutions

Statistical Inference
1. In an urn there is an unknown number N of balls numbered from 1
to N . The number of N should be estimated. A ball from the urn is
used for this purpose and his number is noted. Describe the random
variable X= the number of the drawn ball.

(a) Determine the distribution of X depending on N . Calculate the

expected value and variance of X.
(b) Show that T (X) = 2X − 1 is an unbiased estimator for N is.
(c) Calculate for N = 4 and N = 5 the probability for N to be exactly
estimated at T .
(d) Calculate the variance of T .

Answer:
1
(a) uniform distribution: P (X = k) = ϑ
for k = 1, ..., θ, i.e. E(X) =
2
N +1
2
, Var(X) = N12−1
(b) E(T (X)) = E(2X − 1) = 2E(X) − 1 = N
1 N +1

N +1 ∈N
(c) P (T (X) = N ) = P (2X−1 = N ) = P (X = 2
) = N 2 ⇒
0 else
N=4: P (T (X) = N ) = 0 and N=5: P (T (X) = N ) = 1/5
N 2 −1
(d) V ar(T ) = V ar(2X − 1) = 4V ar(X) = 3

2. Fish are caught from a lake, until you get n (n ≥ 3) fishes of a certain
species A. The random variable X describe the number of all caught
fishes to this time. The lake contained a great number of fishes, so
that it can be assumed that the ratio p of the number of fishes of the
species A to the total number of all fish of the lake does not change,
when some fish are caught out of the lake.
k−1 n


(a) Show that Pp (X = k) = n−1 p (1 − p)k−n , k = n, n + 1, . . .

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 2
n−1
(b) Show that T (X) = is an unbiased estimator for p.
X −1
Answer:

(a) X ∈ {n, n + 1, n + 2, ....}

P (X = k) = P ({n-1 fishes from species A are among the first k-1 caught fishes} ∪
{k th caught fish is a fish from species A})
 
k − 1 n−1
= p (1 − p)k−1−n+1 · p
n−1
 
k−1 n
= p (1 − p)k−n
n−1
(b)
n−1
E(T (X)) = E( )
X −1
∞  
X n−1 k−1 n
= · p (1 − p)k−n
k=n
k−1 n−1
∞  
X k−2 n
= p (1 − p)k−n
k=n
n − 2
∞  
X k − 2 n−1
= p· p (1 − p)k−n
k=n
n − 2
∞  
X k−1
= p· pn−1 (1 − p)k+1−n
k=n−1
(n − 1) − 1
∞
X
= p· Pp (X̃ = k) = p
k=n−1

with X̃ number of all caught fishes until n-1 fishes of a certain are
get.

Maximum Likelihood Estimation

1. A ticket inspector checks for Frankfurt S-Bahn lines the tickets from
the passengers. He keeps checking until he sees a passenger without
valid ticket. He then collects the increased fare and starts after a break
with a new check of the tickets.

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 3
For 10 such check runs, he shall have

42 50 40 64 30 36 68 42 46 48

until he have found a non valid ticket.

Determine a maximum likelihood estimator based on the given numbers
for p share of nonvalid tickets among all checked ticktes.
Answer: ϑ ∈ (0, 1) = ratio non valid tickets
The random variable X = “number of tickets until the first non valid
ticket” is geometricaly distributed with parameter ϑ, i.e. P (X = k) =
(1 − ϑ)k−1 ϑ, k = 1, 2, ...
Likelihoodfunction
n
Y Pn
L(x1 , ..., xn ; ϑ) = (1 − ϑ)xi −1 ϑ = ϑn (1 − ϑ)( i=1 xi )−n
i=1

Easier to consider is P
f (ϑ) = ln L(x1 , ..., xnP; ϑ) = n ln ϑ + ( i=1 nxi − n) ln(1 − ϑ)
n
xi −n
From f 0 (ϑ) = nϑ − i=1 1−ϑ
= 0 we get, ϑ̂ = Pnn xi . f 0 has a sign
i=1
change from + to -. Thus there is local maximum.
Here: ϑ̂ = 0.0215

2. A device consists of the components K1 , K2 and K3 . The device be-

comes defective as soon as one or more of the components is defective.
The lifetimes L1 , L2 and L3 (in h) of the three components are inde-
pendent random variables.
1 − e−λx für x ≥ 0

The distribution function of L1 is F1 (x) =
0 sonst
 √
3
1 − e−λ x für x > 0
The distribution functions of L2 and L3 are F2 (x) = .
0 sonst
λ is an unknown parameter > 0.

(a) Calculate the distribution function and density for the lifetime S
of the device.
(b) When measuring the lifetime of randomly from production of the
devices removed resulted in following values in hours:

82.2 94.0 122.5 95.8 106.4

Use a maximum likelihood estimator to determine the an estimate

for λ.

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 4
Answer:

(a)

P (S ≤ s) = 1 − P (S > s) = 1 − P (S1 > s) · P (S2 > s) · P (S3 > s)

 √
3
1 − e−λ(s+2 s) für s > 0
=
0 sonst
√
3
density function: f (λ, s) = λ(1 + 3
2
√
3
s2
)e−λ(s+2 s)

(b) Likelihoodfunktion
5
5
Y 2 √
−λ(si +2 3 si )
L(s1 , ..., s5 ; λ) = λ (1 + p )e
i=5 3 3 s2i

Taking the logarithm of the likelihood we get

4
!
X 2 √
f (λ) = ln(L(s1 , ..., s5 ; λ) = 5 ln λ+ ln(1 + p
3 2
) − λ(si + 2 3 si )
i=1 3 si

Taking the first derivative of f (λ) and set it zero

5
0 5 X √
f (λ) = − (si + 3 si ) = 0
λ i=1

we get that we have a local maximum at

5
λ̂ = P5 √ = 0.00914
i=1 (si + 2 si )
3

3. To determine the number of N of red deers living in a precinct region

7 red deer were caught and marked in a trapping action. Afterwards
the animals were again released. After a certain time, another trapping
action was started. Thereby 3 red deer were caught, whereby 2 already
were marked. It is assumed that between is no influx or outflow of red
deer in the region and that the animals were able to pass the region
within a short period of time.

(a) Determine a maximum likelihood estimator for the total number

N of the red deer living in the region.
(b) A third trapping action started, where 8 red deers were caught.
4 of them were marked. What is no the maximum likelihood
estimation of N ?

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 5
Answer:

(a) If N denotes the unknown number of red deers and X denotes the
random variables which counts the number of caught marked red
deers in the second trapping action we have
7 N −7



2
PN (X = 2) = N

1
3

The Likelihoodfunktion L(2; N ) is nothing else then this proba-

bility.

Likelihod function

N p
0.5

7 0.00
prob. of the observation

0.4

8 0.38
9 0.50
0.3

10 0.52
11 0.51
0.2

12 0.48
0.1

13 0.44
14 0.40
0.0

15 0.37
16 0.34 10 20 30 40 50

N
blue = first caught, red: first and second caught

⇒ maximum likelihood estimation of N is 10.

(b) Let Y denotes the number of caught marked red deers in the third
trapping action
7+1 N −7−1



4
PN (Y = 4) = N

4
8

The probability of both observation is PN (X = 2) · PN (Y = 4),

which is the likelihood function L(2, 4; N )

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 6
N p
9 0
10 0
11 0
12 0.0675
13 0.120
14 0.141
15 0.141
16 0.128
17 0.112
18 0.0951
19 0.0795
20 0.0659
⇒ maximum likelihood estimation of N is 14.

Confidence Intervals
1. A population is known to be normally distributed with a standard
deviation of 2.8.

(a) Compute the 95% confidence interval on the mean based on the
following sample of nine: 8, 9, 10, 13, 14, 16, 17, 20, 21.
(b) Now compute the 99% confidence interval using the same data.

Answer: Assumption: Normal distribution with known standard de-

viation σ = 2.8

(a) Wanted: 95% confidence interval for µ

n = 9
Data: P9 xi 128
x̄ = i=1 9 = 9 = 14.22
Confidence interval:
2.8
x̄ ± √σn · u0.975 = 14.22 ± √ 9
· 1.96 = 14.22 ± 1.829 ⇒ [12.39, 16.05]
(b) Wanted: 99% confidence interval for µ
n = 9
Data: P9 xi 128
x̄ = i=1 9 = 9 = 14.22
Confidence interval: √
x̄ ± √σn · u0.995 = 14.22 ± √2.8
9
· 2.5758 = 14.22 ± 7.2122/ 9 ⇒
[11.82, 16.62]

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 7
##################################################
# A p o p u l a t i o n i s known t o be n o r m a l l y d i s t r i b u t e d
# with a standard d e v i a t i o n o f 2 . 8 .
#
# f i l e : i n f s t a t c o n f i n t e r v a l n o r m a l m e a n .R
##################################################

# a ) Compute t h e 95% c o n f i d e n c e i n t e r v a l on t h e mean

s a m p l e <= c ( 8 , 9 , 1 0 , 1 3 , 1 4 , 1 6 , 1 7 , 2 0 , 2 1 )
a l p h a <= 0 . 0 5
m <= mean ( s a m p l e )
m
s <= 2 . 8
q a <= qnorm(1 = a l p h a / 2 , 0 , 1 )
q a
u <= m=q a * s / s q r t ( l e n g t h ( s a m p l e ) )
o <= m+q a * s / s q r t ( l e n g t h ( s a m p l e ) )
u;o

# b ) Now compute t h e 99% c o n f i d e n c e interval using t h e same d a t a .

a l p h a <= 0 . 0 1
q a <= qnorm(1 = a l p h a / 2 , 0 , 1 )
q a
u <= m=q a * s / s q r t ( l e n g t h ( s a m p l e ) )
o <= m+q a * s / s q r t ( l e n g t h ( s a m p l e ) )
u;o

# S o l u t i o n a p p l y i n g z . t e s t ( ) from t h e TeachingDemos p a c k a g e
l i b r a r y ( TeachingDemos )
z . t e s t ( x= sample , s d = 2 . 8 , a l t e r n a t i v e = ” two . s i d e d ” , c o n f . l e v e l = 0 . 9 5 ) $ c o n f . i n t
# a)
z . t e s t ( x = sample , s d = 2 . 8 , a l t e r n a t i v e = ” two . s i d e d ” , c o n f . l e v e l = 0 . 9 9 ) $ c o n f . i n t # b )

2. You take a sample of 22 from a population of test scores, and the mean
of your sample is 60.

(a) You know the standard deviation of the population is 10. What
is the 99% confidence interval on the population mean?
(b) Now assume that you do not know the population standard devi-
ation, but the standard deviation in your sample is 10. What is
the 99% confidence interval on the mean now?

Hint: Assume that the test scores follow a normal distribu-

tion.
n = 22
Answer: Assumption: Normal distribution, Data:
x̄ = 60

(a) Wanted: 99% confidence interval for µ

Assumption: Known standard deviation σ = 10
Confidence interval:
x̄ ± √σn · u0.995 = 60 ± √1022 · 2.5758 = 60 ± 5.492 ⇒ [54.508, 65.492]
(b) Wanted: 99% confidence interval for µ
Assumption: Unknown standard deviation, but already estimated
s = 10 (i.e. tn−1 −distribution is used)
Confidence interval:
x̄ ± √σn · t21,0.995 = 60 ± √1022 · 2.8314 = 60 ± 6.036 ⇒ [53.963, 66.036]

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 8
##################################################
# You t a k e a s a m p l e o f 22 from a p o p u l a t i o n o f t e s t
# s c o r e s , and t h e mean o f y o u r s a m p l e i s 6 0 .
#
# f i l e : i n f s t a t c o n f i n t e r v a l l n o r m a l m e a n s d u n k n o w n .R
###################################################
n <= 22
m <= 60

# a ) You know t h e s t a n d a r d d e v i a t i o n o f t h e p o p u l a t i o n i s 1 0 . What

# i s t h e 99\% c o n f i d e n c e i n t e r v a l on t h e p o p u l a t i o n mean .
a l p h a <= 0 . 0 1
s <= 10
q a <= qnorm(1 = a l p h a / 2 , 0 , 1 )
q a
u <= m=q a * s / s q r t ( n )
o <= m+q a * s / s q r t ( n )
u;o

# S o l u t i o n a p p l y i n g z . t e s t ( ) from t h e TeachingDemos p a c k a g e
l i b r a r y ( TeachingDemos )
z . t e s t ( x = m, s d = 1 0 , a l t e r n a t i v e = ” two . s i d e d ” , n = 2 2 , c o n f . l e v e l = 0 . 9 9 ) $conf . int
# b ) Now assume t h a t you do n o t know t h e p o p u l a t i o n s t a n d a r d
# d e v i a t i o n , but t h e s t a n d a r d d e v i a t i o n i n y o u r s a m p l e i s 1 0 . What
# i s t h e 99\% c o n f i d e n c e i n t e r v a l on t h e mean now?
s s a m p l e <= 10
t a <= q t (1 = a l p h a / 2 , n = 1)
t a
u <= m= t a * s / s q r t ( n )
o <= m+t a * s / s q r t ( n )
u;o

3. Calculate for the below given sample from a normally distributed pop-
ulation the 95% confidence intervals

(a) for the mean, if the standard deviation is 2

(b) for the mean, if the standard deviation is unknown
(c) for the variance, if the mean is 250
(d) for the variance, if the mean is unknown

xi : 247.4, 249.0, 248.5, 247.5, 250.6, 252.2, 253.4, 248.3, 251.4, 246.9,
249.8, 250.6, 252.7, 250.6, 250.6, 252.5, 249.4, 250.6, 247.0, 249.4

Answer: sample size n=20, x̄ = 249.92, s = 1.9479, α = 0.05

h i
(a) x̄ − u1−α/2 √σn , x̄ + u1−α/2 √σn = [249.04, 250.80]
h i
s s
(b) x̄ − tn−1,1−α/2 n , x̄ + tn−1,1−α/2 n = [229.01, 250.83
√ √
 
Qn Qn
= [2.11, 7.53] with Qn = ni=1 (xi − µ)2 = 72.22
P
(c) χ2 , χ2
n,1−α/2 n,α/2
 
(n−1)s2 (n−1) s2
(d) χ2 , χ2 = [2.19, 8.09]
n−1,1−α/2 n−1,α/2

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 9
####################################################
# C a l c u l a t e f o r t h e g i v e n s a m p l e from n o r m a l l y
# d i s t r i b u t e d p o p u l a t i o n t h e 95% c o n f i d e n c e i n t e r v a l s
# a ) f o r t h e mean , i f t h e s t a n d a r d d e v i a t i o n i s 2
# b ) f o r t h e mean , i f t h e s t a n d a r d d e v i a t i o n i s unknown
# a ) f o r t h e v a r i a n c e , i f t h e mean i s 250
# a ) f o r t h e v a r i a n c e , i f t h e mean i s unknown
#
# f i l e : i n f s t a t c o n f i n t e r v a l l n o r m a l m u s i g m a .R
#####################################################

# c r e a t e sample v a l u e s
# s . v a l u e s <= round ( rnorm ( n=20 , mean = 2 5 1 , s d = 2 ) , 1 )
s . v a l u e s <= c ( 2 4 7 . 4 , 2 4 9 . 0 , 2 4 8 . 5 , 2 4 7 . 5 , 2 5 0 . 6 , 2 5 2 . 2 , 2 5 3 . 4 , 2 4 8 . 3 , 2 5 1 . 4 , 2 4 6 . 9 ,
249.8 ,250.6 ,252.7 ,250.6 ,250.6 ,252.5 ,249.4 ,250.6 ,247.0 ,249.4)
# c h a r a c t e r i s t i c s o f the sample
n <= l e n g t h ( s . v a l u e s )
x b a r <= mean ( s . v a l u e s )
s <= s d ( s . v a l u e s )
# l e v e l 1= a l p h a
a l p h a <= 0 . 0 5

# c o n f i d e n c e i n t e r v a l l s f o r mu
# a ) a s s u m p t i o n : s i gm a = 2
s i gm a <= 2
l . a <= x b a r = qnorm(1 = a l p h a / 2 ) * s ig m a / s q r t ( n )
u . a <= x b a r + qnorm(1 = a l p h a / 2 ) * s ig m a / s q r t ( n )
l .a; u.a
# b ) a s s u m p t i o n : s i g ma = unknown
l . b <= x b a r = q t (1 = a l p h a / 2 , d f = n = 1) * s / s q r t ( n )
u . b <= x b a r + q t (1 = a l p h a / 2 , d f = n = 1) * s / s q r t ( n )
l .b; u.b

# c o n f i d e n c e i n t e r v a l l s f o r s i gm a ˆ2
# c ) a s s u m p t i o n : mu = 250
mu <= 250
Qn <= sum ( ( s . v a l u e s = mu) ˆ 2 )
l . c <= Qn/ q c h i s q (1 = a l p h a / 2 , d f = n )
u . c <= Qn/ q c h i s q ( a l p h a / 2 , d f = n )
l .c; u. c
# d ) a s s u m p t i o n : mu unknown
l . d <= ( n = 1) * s ˆ2/ q c h i s q (1 = a l p h a / 2 , d f = n = 1)
u . d <= ( n = 1) * s ˆ2/ q c h i s q ( a l p h a / 2 , d f = n = 1)
l .d; u.d

# s o l u t i o n s a p p l y i n g z . t e s t ( ) , s i gm a . t e s t ( ) from TeachingDemos and t . t e s t ( )

l i b r a r y ( TeachingDemos )
z . t e s t ( x = s . v a l u e s , sd = 2 , a l t e r n a t i v e = ” two . s i d e d ” , c o n f . l e v e l = 0 . 9 5 ) $ c o n f . i n t # a )
t . t e s t ( x = s . v a l u e s , a l t e r n a t i v e = ” two . s i d e d ” , c o n f . l e v e l = 0 . 9 5 ) $ c o n f . i n t
# b)
s i gm a . t e s t ( x = s . v a l u e s , a l t e r n a t i v e = ” two . s i d e d ” , c o n f . l e v e l = 0 . 9 5 )
# d)

4. At a telemarketing firm, the length of a telephone solicitation (in sec-

onds) is a normally distributed random variable with mean µ and stan-
dard deviation σ, both unknown. A sample of 51 calls has mean length
300 and standard deviation 60.

(a) Construct the 95% confidence upper bound for µ.

(b) Construct the 95% confidence lower bound for σ.

Answer: Sample size n = 51 and sample mean x̄ = 300 and sample

standard deviation s = 60

(a) Wanted: Confidence interval for µ at level 1 − α = 95%

In
h general we have the two-sided confidence i interval.
x̄ − tn−1, 1− α2 · √sn , x̄ + tn−1, 1− α2 · √sn

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 10
A
 one-sided confidence interval
i h (upper boundary): i
−∞, x̄ + tn−1, 1−α · √n = −∞, 300 + 1.6759 · √6051 = (−∞, 314, 23]
s

with t50,0.95 = 1.6759

(b) Wanted: Confidence interval for σ at level 1 − α = 95%  
2 (n−1)s2
In general we have the two sided confidence interval for σ 2 : χ2(n−1)s , χ2n−1,α/2
n−1,1−α/2

A
hqone-sided confidence
 interval (lower boundary) for σ:
(n−1)s2
χ2
, ∞ = [51, 57 ; ∞) with χ2n−1,1−α = χ250,0.95 = 67.505
n−1,1−α

####################################################
# At a t e l e m a r k e t i n g f i r m , t h e l e n g t h o f a t e l e p h o n e
# s o l i c i t a t i o n ( in seconds ) i s a normally d i s t r i b u t e d
# random v a r i a b l e w i t h mean mu and s t a n d a r d d e v i a t i o n
# sigma , both unknown . A s a m p l e o f 50 c a l l s h a s mean
# l e n g t h 300 and s t a n d a r d d e v i a t i o n 6 0 .
#
# f i l e : i n f s t a t c o n f i n t e r v a l t e l e f i r m .R
#####################################################
n <= 5 0 ; m <= 3 0 0 ; s s a m p l e <= 6 0 ; a l p h a <= 0 . 0 5

# a ) C o n s t r u c t t h e 95% c o n f i d e n c e u p p e r bound f o r mu .
t a <= q t (1 = a l p h a , n = 1)
t a
o <= m+t a * s s a m p l e / s q r t ( n )
o

# b ) C o n s t r u c t t h e 95% c o n f i d e n c e l o w e r bound f o r s ig m a .
c h i <= q c h i s q (1 = a l p h a , n = 1)
chi
u <= ( n = 1) * s s a m p l e ˆ2/ c h i
sqrt (u)

5. At a certain farm the weight of a peach (in ounces) at harvest time

is a normally distributed random variable with standard deviation 0.5.
How many peaches must be sampled to estimate the mean weight with
a margin of error ±0.2 and with 95% confidence.
Answer: Standard deviation σ = 0.5 known.
Confidence interval: x̄ ± √σn · u0.975
√
Wanted: n with √σn ·u0.975 = 0.2 i.e. √0.5
n
·1.96 = 0.2 i.e. n = 0.5·1.96
0.2
i.e.
0.5·1.96 2


n= 0.2
= 4.92 = 24.01 ⇒ n = 25 since we must round upwards.
########################################################
# At a c e r t a i n farm t h e w e i g h t o f a p e a c h ( i n o u n c e s )
# a t h a r v e s t t i m e i s a n o r m a l l y d i s t r i b u t e d random
# v a r i a b l e w i t h s t a n d a r d d e v i a t i o n 0 . 5 . How many p e a c h e s
# must be sampled t o e s t i m a t e t h e mean w e i g h t w i t h a
# margin o f e r r o r pm 0 . 2 and w i t h 95% c o n f i d e n c e .
#
# f i l e : i n f s t a t c o n f i n t e r v a l p e a c h .R
########################################################

a l p h a <= 0 . 0 5 ; s <= 0 . 5 ; margin <= 0 . 2

q a <= qnorm(1 = a l p h a / 2 , 0 , 1 ) ; q a
n <= c e i l i n g ( ( q a * s / margin ) ˆ 2 )
n

6. You read about a survey in a newspaper and find that 70% of the 250
people sampled prefer candidate A.

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 11
(a) Compute the 95% confidence interval.
(b) You are surprised by this survey because you thought that more
like 50% of the population preferred this candidate. Based on this
sample, is 50% a possible population proportion?
n = 250
Answer: Data:
x̂ = 0.70

(a) Wanted: Confidence

q interval for p at level 1 − α = 0.95
p̂ ± u1− α2 · p̂(1−p̂)
n
q
We have: 0.70 ± 1.96 · 0.70(1−0.70)
250
= 0.70 ± 0.057
i.e. [0.6432, 0.7568]
(b) Possible, but with a very low probability since 50% is not in the
confidence interval.
########################################################
# You r e a d a b o u t a s u r v e y i n a n e w s p a p e r and f i n d
# t h a t 70% o f t h e 250 p e o p l e sampled p r e f e r C a n d i d a t e A .
# a ) Compute t h e 95% c o n f i d e n c e i n t e r v a l .
# b ) You a r e s u r p r i s e d by t h i s s u r v e y b e c a u s e you t h o u g h t
# t h a t more l i k e 50% o f t h e p o p u l a t i o n p r e f e r r e d t h i s
# c a n d i d a t e . Based on t h i s sample , i s 50% a p o s s i b l e
# population proportion ?
#
# f i l e : i n f s t a t c o n f i n t e r v a l p r o p s u r v e y .R
########################################################

n <= 2 5 0 ; p <= 0 . 7 ; a l p h a <= 0 . 0 5

# normal a p p r o x i m a t i o n
l . appr <= p = qnorm(1 = a l p h a / 2 ) * s q r t ( p * (1 = p ) / n )
u . appr <= p + qnorm(1 = a l p h a / 2 ) * s q r t ( p * (1 = p ) / n )
l . appr ; u . appr

# exact
xp <= s e q ( 0 , 1 , l e n g t h =1+10ˆ4)
l . ex <= xp [ min ( which ( qbinom(1 = a l p h a / 2 , n , xp ) == p * n ) ) ]
u . ex <= xp [ max ( which ( qbinom ( a l p h a / 2 , n , xp ) == p * n ) ) ]
l . ex ; u . ex

# e x a c t c o n f i d e n c e i n t e r v a l w i t h R= f u n c t i o n
binom . t e s t ( x = 0 . 7 * 2 5 0 , n =250 , c o n f . l e v e l =1= a l p h a ) $ c o n f . i n t

7. A researcher was interested in knowing how many people in the city

supported a new tax. He sampled 100 people from the city and found
that 40% of these people supported the tax. What is the upper limit of
the 95% (one-side) confidence interval on the population proportion?
Answer: Survey with n = 100 and 40% approve the taxes
Wanted: Upper-boundary confidence
q interval for a proportion p = at
level 1 − α = 0.95: p̂ + u1−α · p̂(1−p̂)
n
We have n = 100, p̂ q = 0.4 and 1 − α = 0.95 ⇒ u1−α = 1.645, i.e. we
0.4·0.6
become 0.4 + 1.645 · 100
= 0.48

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 12
##################################################
# A r e s e a r c h e r was i n t e r e s t e d i n knowing how many
# p e o p l e i n t h e c i t y s u p p o r t e d a new t a x . She sampled
# 100 p e o p l e from t h e c i t y and f o u n d t h a t 40% o f
# t h e s e p e o p l e s u p p o r t e d t h e t a x . What i s t h e u p p e r
# l i m i t o f t h e 95% ( one = s i d e ) c o n f i d e n c e i n t e r v a l
# on t h e p o p u l a t i o n p r o p o r t i o n ?
#
# f i l e : i n f s t a t c o n f i n t e r v a l l p r o p o n e s i d e d .R
##################################################

n <= 1 0 0 ; p <= 0 . 4 ; a l p h a <= 0 . 0 5

# normal a p p r o x i m a t i o n
u . appr <= p + qnorm(1 = a l p h a ) * s q r t ( p * (1 = p ) / n )
u . appr

# exact
xp <= s e q ( 0 , 1 , l e n g t h =1+10ˆ4)
u . ex <= xp [ max ( which ( qbinom ( a l p h a , n , xp ) == p * n ) ) ]
u . ex

# e x a c t c o n f i d e n c e i n t e r v a l w i t h R= f u n c t i o n
binom . t e s t ( x =40 , n =100 , a l t e r n a t i v e = ” l e s s ” ,
c o n f . l e v e l =1= a l p h a ) $ c o n f . i n t

8. An advertising agency wants to construct a 99% confidence lower bound

for the proportion of dentists who recommend a certain brand of tooth-
paste. The margin of error is to be 0.02. How large should the sample
be?
Answer: The lower boundary at level 1 − α = 0.99 for the proportion
p is denoted z.qThus, we have q
z = p̂ − u1−α · p̂(1−p̂)
n
with u1−α · p̂(1−p̂)
n
≤ 0.02
α = 0.01 ⇒ u0.99 = 2.326 q
p̂(1−p̂)
n is unknown, thus 2.326 · n
≤ 0.02 i.e.
2
2.326 · p̂(1 − p̂) ≤ 0.02 · n i.e. n ≥ 2.326
2 2
0.022
p̂(1 − p̂)
For which p̂ has the function y = p̂(1 − p̂) a maximum? We take the
derivative: y = p̂ − p̂2 ⇒ y 0 = 1 − 2p̂ and then y 0 = 1 − 2p̂ = 0 ⇒ p̂ = 12 .
Thus, we have y = p̂(1 − p̂) ≤ 12 · 21 = 14 .
2
This gives n ≥ 2.326
0.022
· 41 = 3381
2.3262
If we suppose that p ≤ 0.25, i.e. n ≥ 0.022
· 14 · 3
4
= 2537
##################################################
# An a d v e r t i s i n g a g e n c y wants t o c o n s t r u c t a 99%
# c o n f i d e n c e l o w e r bound f o r t h e p r o p o r t i o n o f
# d e n t i s t s who recommend a c e r t a i n brand o f t o o t h p a s t e .
# The margin o f e r r o r i s t o be 0 . 0 2 . How l a r g e s h o u l d
# t h e s a m p l e be ?
#
# f i l e : i n f s t a t c o n f i n t e r v a l p r o p s a m p l e s i z e .R
##################################################

a l p h a <= 0 . 0 1 ; margin <= 0 . 0 2

c <= qnorm(1 = a l p h a , 0 , 1 )
f <= s e q ( 0 , 1 , l e n g t h =101)
n <= max ( c e i l i n g ( c ˆ2 * f * (1 = f ) / ( margin ˆ 2 ) ) )
n

9. The interval [45.6, 47.8] is a symmetric 99% confidence interval for the
unknown parameter µ based on a sample x1 , . . . , x10 from a normal

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 13
2
distribution N (µ, σ ) with unknown σ. Calculate the sample mean x̄
and the sample standard deviation s.
Answer: Mean: x̄ = 45.6+47.8 = 46.7 and using the lower limit 45.6 we
2 √ √
get 45.6 = x̄ − t9, 0.995 · √n i.e. s = tx̄−45.6
s
9, 0.995
· n = 46.7−45.6
3.25
· 10 = 1.07

10. The waiting time at the pay desk of a certain supermarket is normally
distributed with mean waiting time µ and known standard deviation
σ = 1, 8 minutes. A confidence interval for the mean waiting time
(in minutes) for this supermarket is [5.12; 8.32]. If the sample size is
n = 10, what is then the confidence level?
Answer: The length of the interval is 8.32 − 5.12 and
8.32 − 5.12 = 2 · u1− α2 · √σn = 2 · u1− α2 · √1.8
10
i.e. u1− α2 = 2.81 and the
α
normal distribution table gives 1 − 2 = 0.9975 i.e. α ≈ 0.005. So the
confidence level is 1 − α = 99.5%.

11. R programming task: Consider an urn with M white balls and N-

M black. n balls are drawn without replacement and X denotes the
number of white balls in the sample. N=500 and n=50 are known but M
the number of white balls is unknown. Construct an two sided 1 − α =
0.95 confidence intervall for M based on the H(N,M,n)-distribution of
X. Compare it with a binomial and a normal approximation.
Answer:
########################################################
# C o n s i d e r an urn w i t h M w h i t e b a l l s and N=M b l a c k . n
# b a l l s a r e drawn w i t h o u t r e p l a c e m e n t and X d e n o t e s t h e
# number o f w h i t e b a l l s i n t h e s a m p l e . N=500 and n=50
# a r e known but M t h e number o f w h i t e b a l l s i s unknown .
# C o n s t r u c t an two s i d e d 1= a l p h a =0.95 c o n f i d e n c e i n t e r v a l l
# f o r M b a s e d on t h e H(N,M, n)= d i s t r i b u t i o n o f X . Compare
# i t w i t h a b i n o m i a l and a normal a p p r o x i m a t i o n .
#
# f i l e : i n f s t a t c o n f i n t e r v a l h y p e r g e o M .R
########################################################

library ( tidyverse )

# urn m o d e l l : N t o t a l number o f b a l l s , M = number o f white

# b a l l s , n = number o f drawn b a l l s
# X = number o f w h i t e b a l l s ˜ H(N,M, n )
N <= 500
n <= 50
a l p h a <= 0 . 0 5

# s y m m e t r i c i n t e r v a l s [ l b , ub ] f o r X w i t h p r o b a b i l i t y 1= a l p h a
# for d i f f e r e n t values of M
s y . i n t e r v a l s <= t i b b l e (
M = 0 : N,
# q u a n t i l s o f H(N,M, n )
l b = q h y p e r ( a l p h a / 2 ,M, N=M, n ) ,
ub = q h y p e r (1 = a l p h a / 2 ,M, N=M, n )
)
# p l o t of the the i n t e r v a l s
p l o t ( x=s y . i n t e r v a l s $ M , y=s y . i n t e r v a l s $ l b , c o l =” b l u e ” ,
t y p e = ”p ” ,
x l a b = ”M” , y l a b = ” l o w e r and u p p e r bounds ” ,
main = ” s y m m e t r i c 95% i n t e r v a l s f o r X” )
p o i n t s ( x=s y . i n t e r v a l s $ M , y=s y . i n t e r v a l s $ u b , c o l =”r e d ” )

# Mention t h e l b = and ub= f u n c t i o n s a r e n o t s t r i c t l y m o n o t o n o u s l y

# i n c r e a s i n g : u s e f o r g i v e n v a l u e o f X t h e min o f t h e

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21

 14
# c o r r e s p o n d i n g ub v a l u e s and t h e max o f t h e c o r r e s p o n d i n g l b
# v a l u e s o f M a s an i n v e r s e o f t h e two f u n c t i o n . These v a l u e s
# a r e t h e bounds o f t h e c o n f i d e n c e i n t e r v a l s .
ex . c o n f . i n t e r v a l l <= f u n c t i o n ( x ) {
return ( c (
s y . i n t e r v a l s %>%
f i l t e r ( ub == x ) %>%
mutate ( l = min (M) ) %>%
s e l e c t ( l ) %>%
u n i q u e ( ) %>%
as . numeric ( ) ,
s y . i n t e r v a l s %>%
f i l t e r ( l b == x ) %>%
mutate ( u = max (M) ) %>%
s e l e c t ( u ) %>%
u n i q u e ( ) %>%
as . numeric ( )
))
}

# The binom . t e s t ( x , n ) f u n c t i o n r e t u r n s i n t h e v a r i a b l e
# c o n f . i n t t h e c o n f i d e n c e i n t e r v a l f o r p=M/N i f t h e y a r e X
# w h i t e b a l l s i n a s a m p l e o f n b a l l s drawn from t h e urn
# with replacement
binom . appr . c o n f . i n t e r v a l l <= f u n c t i o n ( x ) {
return (
c(
binom . t e s t ( x , n , c o n f . l e v e l = 1= a l p h a ) $ c o n f . i n t [ 1 ] * N,
binom . t e s t ( x , n , c o n f . l e v e l = 1= a l p h a ) $ c o n f . i n t [ 2 ] * N
)
)
}

# normal a p p r o x i m a t i o n o f t h e c o n f i d e n c e i n t e r v a l f o r an
# unknown p r o p o r t i o n i f x w h i t e b a l l s a r e i n a s a m p l e o f
# n b a l l s drwan w i t h r e p l a c e m e n t
normal . appr . c o n f . i n t e r v a l l <= f u n c t i o n ( x ) {
return (
c(
N * ( x /n =qnorm(1 = a l p h a / 2 ) * s q r t ( x * (1 = x /n ) / n ˆ 2 ) ) ,
N * ( x /n +qnorm(1 = a l p h a / 2 ) * s q r t ( x * (1 = x /n ) / n ˆ 2 ) )
)
)
}

# t i b b l e o f t h e bounds o f t h e c o n f i d e n c e i n t e r v a l l s for M
# for a l l possibloe values of X
t a b <= t i b b l e (
X = 0 : n ) %>%
g r o u p b y (X) %>%
mutate ( ex . l b=ex . c o n f . i n t e r v a l l (X ) [ 1 ] ,
ex . ub=ex . c o n f . i n t e r v a l l (X ) [ 2 ] ,
binom . l b=binom . appr . c o n f . i n t e r v a l l (X ) [ 1 ] ,
binom . ub=binom . appr . c o n f . i n t e r v a l l (X ) [ 2 ] ,
norm . l b=normal . appr . c o n f . i n t e r v a l l (X ) [ 1 ] ,
norm . ub=normal . appr . c o n f . i n t e r v a l l (X ) [ 2 ]
)

# p l o t o f a l l bounds
p l o t ( x=tab $ X , y=t a b $ e x . l b , c o l =”r e d ” ,
x l a b = ” x ” , y l a b = ”M” ,
main = ”95% c o n f i d e n c e i n t e r v a l l f o r M i n H(N=500 ,M, n =50)” ,
sub = ” r e d = e x a c t , b l u e = b i n o m i a l approx , b l a c k = normal approx . ” )
p o i n t s ( x=tab $ X , y=t a b $ e x . ub , c o l =”r e d ” )
p o i n t s ( x=tab $ X , y=tab $ binom . l b , c o l =” b l u e ” )
p o i n t s ( x=tab $ X , y=tab $ binom . ub , c o l =” b l u e ” )
p o i n t s ( x=tab $ X , y=tab $ norm . l b , c o l =” b l a c k ” )
p o i n t s ( x=tab $ X , y=tab $ norm . ub , c o l =” b l a c k ” )

Prof. Dr. Falkenberg, Faculty 2 Statistics WS 20/21