MATH10282

Two hours

Statistical tables to be provided

THE UNIVERSITY OF MANCHESTER

INTRODUCTION TO STATISTICS

30 May 2017
14.00 – 16.00

Answer ALL FOUR questions in Section A (10 marks each) and TWO of the THREE questions
in Section B (20 marks each). If more than TWO questions from Section B are attempted, then
credit will be given for the best TWO answers.

Electronic calculators may be used, provided that they cannot store text.

1 of 8 P.T.O.
 MATH10282
SECTION A

Answer ALL four questions

A1. Suppose that we have a sample of observations x1 , . . . , xn obtained by random sampling from a
continuous distribution with probability density function f (x).

We wish to calculate a density histogram. Suppose that x1 , . . . , xn ∈ (a1 , aK+1 ] and that the range
(a1 , aK+1 ] is divided evenly into K subintervals, or bins, Bk = (ak , ak+1 ], k = 1, . . . , K, of width h.

(i) Write down an expression for the function Hist(x) defining the density histogram based on bins
Bk . Explain any notation you use.
[2 marks]

One possible way of estimating the distribution mean,

Z ∞
µ= xf (x)dx ,
−∞

is to substitute Hist(x) as an estimate of the p.d.f. f (x), i.e.

Z ∞
µ̂Hist = x Hist(x)dx .
−∞

(ii) Evaluate this integral to obtain an expression for µ̂Hist .

[5 marks]

A data set is obtained which has the following frequency table:

Interval (10,20] (20,30] (30,40] (40,50] (50,60] (60,70]
Frequency 7 19 38 23 12 1

(iii) Using your answer to part (ii), or otherwise, estimate the mean of the distribution that gener-
ated these data.
[3 marks]

[Total 10 marks]

2 of 8 P.T.O.
 MATH10282

1
Pn
A2. Suppose that X1 , . . . , Xn ∼ N (µ, σ 2 ) independently, and let S 2 = n−1 i=1 (Xi − X̄)2 denote
the sample variance.

(i) Show that an equivalent expression for the sample variance is

n
!
2 1 X
2 2
S = X − nX̄
n − 1 i=1 i

[5 marks]

(ii) State the distribution of (n − 1)S 2 /σ 2 , and give formulae for the end-points of a 100(1 − α)%
confidence interval for σ 2 .
[3 marks]
P10 P10
(iii) A data set is obtained with n = 10, i=1 xi = 859, and i=1 x2i = 74227. Calculate a 95%
confidence interval for σ 2 .
[2 marks]

[Total 10 marks]

3 of 8 P.T.O.
 MATH10282

A3. Let X1 , . . . , Xn be a random sample from U θ − 21 , θ + 12 , where θ is unknown.

 

(i) Show that the estimator

n
1X
θ̂ = X̄n = Xi
n i=1
is an unbiased estimator of θ with variance 1/(12n).
[4 marks]

(ii) If n is large, what is the approximate sampling distribution of X̄? State any results you use.

[2 marks]

(iii) Suppose that a sample of size n = 75 is obtained, for which ni=1 xi = 1147. Use your answer
P
to part (ii) to estimate the approximate probability that, for a future sample of size m = 80
from the same distribution, the value of X̄m is greater than 15.25.
[4 marks]

[Total 10 marks]

4 of 8 P.T.O.
 MATH10282

A4. Suppose that X1 , . . . , Xn ∼ N (µ1 , σ 2 ) independently and Y1 , . . . , Ym ∼ N (µ2 , σ 2 ) independently

of each other and of X1 , . . . , Xn , with µ1 , µ2 and σ 2 all unknown. Let θ = µ1 − µ2 denote the
difference of the two population means. It is desired to test the null hypothesis H0 : θ = θ0 versus
the two-sided alternative hypothesis H1 : θ 6= θ0 .

(i) Write down an appropriate unbiased estimator, σ̂ 2 , of the common variance σ 2 . Write down a
suitably scaled version of σ̂ 2 and state its sampling distribution.
[3 marks]

(ii) Suggest a suitable test statistic for testing H0 vs H1 , and state its exact distribution in the
case that H0 is true. Give an appropriate rejection region to achieve significance level α.

[3 marks]

(iii) A data set is obtained satisfying

n = 10 x̄ = 23.7 s21 = 1.25

m = 12 ȳ = 21.6 s22 = 1.16

Do you reject H0 if θ0 = 1 and α = 0.05? How about if α = 0.01? Show your working.

[4 marks]

[Total 10 marks]

5 of 8 P.T.O.
 MATH10282
SECTION B

Answer TWO of the three questions

B5. Suppose that X1 , . . . , Xn ∼ N (µ, σ 2 ) independently, with µ unknown and σ 2 known. It is of

interest to test the null hypothesis H0 : µ = µ0 versus the one-sided alternative H1 : µ > µ0 at
significance level α.

(i) Define the following terms: Type I error, Type II error, and the significance level of the test
(also known as the size).
[3 marks]

(ii) Write down an appropriate test statistic and rejection region, making sure to define any notation
used. Show that this choice does indeed achieve a significance level of α.
[4 marks]

(iii) Suppose for this part that µ0 = 10, σ 2 = 1 and a data set is obtained with n = 10 and ni=1 xi =
P
106. Do you reject H0 at significance level α = 0.05? What if α = 0.01?
[3 marks]
Pn
(iv) Suppose for this part that a data set is obtained with i=1 xi = nµ0 + 0.2nσ. What is the
minimum value of n for which H0 is rejected at (i) a 5% significance level, and (ii) a 1%
significance level?
[3 marks]

(v) Show that for µ > µ0 the probability of correctly rejecting H0 is equal to
 
µ0 − µ
1 − Φ zα + √ .
σ/ n

Compute the probability of rejecting H0 if µ0 = 10, µ = 10.5, σ = 1, α = 0.05 and n = 10.

[7 marks]

[Total 20 marks]

6 of 8 P.T.O.
 MATH10282

B6. A random variable X has a negative binomial distribution NegBi(r, p), with p ∈ (0, 1) and r a
positive integer, if it has probability mass function (p.m.f.)
 
x+r−1
pX (x) = P(X = x) = (1 − p)r px , x = 0, 1, 2, . . . ,
x
m m!


where above the notation k
= k!(m−k)!
denotes a binomial coefficient.

(i) Show that if X1 , . . . , Xn ∼ NegBi(r, p) independently, with r known and p unknown, then the
likelihood function is
" n  #
Y Xi + r − 1 Pn
L(p) = (1 − p)nr p i=1 Xi .
i=1
Xi
[2 marks]

(ii) Hence show that the maximum likelihood estimator of p is

X̄
p̂ = .
r + X̄
[7 marks]

(iii) Show that  

ar br
P(a < p̂ < b) = P < X̄ < .
1−a 1−b
[6 marks]

(iv) Use the result given in part (iii) to calculate the approximate probability that 0.45 < p̂ < 0.55
when p = 0.5, r = 3 and n = 100.

[Hint: you may use without proof that E(X) = pr/(1 − p) and Var(X) = pr/(1 − p)2 .]

[5 marks]

[Total 20 marks]

7 of 8 P.T.O.
 MATH10282

B7. The population proportion of batteries of Type 1 that are defective is p1 , and the population
proportion of batteries of Type 2 that are defective is p2 . Both p1 and p2 are unknown, and it is
desired to estimate the difference θ = p1 − p2 .

To investigate this, an experimenter will examine n batteries of Type 1 and m batteries of Type 2,
and record how many of each type are defective.

Let X1 denote the number of defectives in the sample of Type 1 batteries, and let X2 denote the
number of defectives in the sample of Type 2 batteries.

(a) (i) Suggest an appropriate estimator, θ̂, for θ and define any notation used.
[2 marks]
(ii) Show that θ̂ is unbiased, and derive an expression for v(n, m) = Var θ̂.
[6 marks]

(b) The investigator considers two possible experimental designs with total sample size n+m = 300:

Design 1: n = 150, m = 150

Design 2: n = 100, m = 200

(i) Show that Var θ̂ is smaller under Design 1 than under Design 2 if and only if

p1 (1 − p1 ) 1
≥ .
p2 (1 − p2 ) 2
[4 marks]
(ii) Hence argue that, when p1 = 0.1, the estimator θ̂ performs better if we use Design 1 rather
than Design 2 provided that p2 < 0.235 or p2 > 0.765.
[4 marks]

(c) Now suppose that the experimenter finds that 13 out of n = 150 batteries of Type 1 are
defective and 17 out of m = 150 batteries of Type 2 are defective.

(i) Calculate an approximate 95% confidence interval for θ.

[4 marks]

[Total 20 marks]

END OF EXAMINATION PAPER

8 of 8