You are on page 1of 5

The University of Western Australia

GENG5507 Risk, Reliability and Safety Planning

Solutions to Tutorial Sheet 5
(Students should take this Tutorial Sheet to the Tutorial in Week 6)

Background material
This weeks questions concern the sampling distribution of the sample mean of a simple random sample.
Required Normal & T distribution tables are given in LMS.
Revision: properties of the Normal distribution
The family of Normal distributions provides accurate probability models for many real-world phenomena.
The Normal distributions also have nice properties. As a result, they often provide a convenient choice
of model.
Linear combinations of Normal random variables are Normal
Suppose that
X N(, 2 ).
Then for any two constants a 6= 0 and b,
aX + b N(a + b, a2 2 ).
As an important example, consider the standardisation of X:
N(0, 1).
Z =

This result applies for all Normal random variables.

Sums of Normal random variables are Normal
If X1 , . . . , Xn are jointly Normally distributed, then the sum
Xi = X1 + + Xn

is also Normally distributed. In particular, if the Xi are iid N(, 2 ), then


Xi N(n, n 2 ).


Sampling distribution of the standardised sample mean from a Normal population

It follows, as an application of the above two results, that the sample mean X = n1 ni=1 Xi of a simple
random sample from a N(, 2 ) population satisfies
X N(, 2 /n),
or equivalently,

/ n

N(0, 1).

Tutorial Questions
1. [Know your Z-tables] If Z N (0, 1) evaluate
(a) P (Z < 1.96); (b) P (Z > 2.13); (c) P (|Z| < 0.05), (d) z such that P (Z < z) = 0.95.

(a) 0.975; (b) 0.166; (c) 0.0398; (d) 1.645.
2. Carbide drill tips are used in drilling oil wells. The life of a carbide tip can be viewed in terms
of the drilling depth (in suitable units) achieved between replacement. In order to estimate the
mean drilling depth of a particular type of carbide tip, the drilling depth lifetimes of 36 carbide
tips are obtained. Assume a simple random sample, i.e. the drilling depths are iid, and assume that
the drilling depths are Normal random variables with unknown mean but with known standard
deviation = 12.
(a) State the distribution of X, in terms of .
(b) Find the probability of the sample mean drilling depth being within 4 units of the true mean.
(c) In order to be more confident that the sample mean is close to the true mean, we may use a
larger sample. Suppose we wish to obtain a sample for which the sample mean drilling depth
has probability 0.99 of being within 4 of the true mean. What is the minimum number of
carbide tips needed to achieve this?

(a) Since this is a simple random sample from a Normal population, X N(, 2 /n). In this
case = 12 and n = 36, so X N(, 122 /36), i.e. X N(, 4).
(b) The required probability is



P | X | 4 = P


= P |Z | 2
(where Z N (0, 1))
= 0.9544


(c) The requirement

| 4
P |X

= 0.99

is equivalent to

P |Z |
/ n

= 0.99,

which implies, using z0.005 = 2.576, that


/ n

= 2.576.

Rearranging, and using = 12, we get

n = 2.576


n = 2.5762

= 59.72.

Thus, we need a sample of size n 60 to meet the requirement.

Revision: Sampling distribution of the standardised sample mean from a

Normal population with unknown variance
Suppose that X is the sample mean of a simple random sample from a N(, 2 ) population. Then,
as noted above,

/ n

N(0, 1).

However, if we do not know , we cannot use this result directly to calculate probabilities for X,
or conduct inference for . But we can estimate , for example, by the sample standard deviation
u 1 X

2 .
Xi2 nX
S = t

It can be shown that the resulting version of the standardised sample mean satisfies

S/ n

tn1 .

The extra uncertainty that results from estimating by S is captured by the thicker tails of the t
distributions (compared to the N(0, 1) distribution).
t distributions for large degrees of freedom
In the limit as n , the t distributions with n 1 degrees of freedom converge to the N(0, 1)
distribution. This means that for large n (a few hundred), we can use the approximation

S/ n

N(0, 1).

3. [Know your T-tables] If T tn evaluate

(a) P (T < 2.262), when n = 9; (b) P (T > 1.734), when n = 18; (c) P (|T | < 2.807) when n = 23;
(d) t such that P (T < t) = 0.025, when n = 29.

(a) 0.975; (b) 0.05; (c) 0.99; (d) -2.045.

4. In a particular manufacturing process, a hole-punch machine is used to punch a hole of a specified

diameter in a strip of metal. To study the reliability of this process, we randomly select n strips of
metal and measure the hole diameters. It can be assumed that the hole diameters X1 , . . . , Xn are
iid Normal random variables, with unknown mean and unknown variance 2 .
(a) State the probability distribution of X = n1 ni=1 Xi .

(b) State the probability distribution of (X )/(S/ n).

(c) For each of (a) and (b), state whether the distribution is exact or an approximation. Give
(a) Since this is a simple random sample from a Normal population, X N(, 2 /n). However,
as is not known, this result cannot be used directly to calculate probabilities pertaining to
X or to draw inferences for based on an observation of X.

(b) Since (X )/(/ n) N(0, 1), it follows immediately that (X )/(S/ n) tn1 . (Here
S is the sample standard deviation.)
(c) Both distributions are exact, since the sample is taken from a Normal population.

Revision: the Central Limit Theorem

Suppose that X1 , . . . , Xn are iid random variables with mean and variance 2 . The Central
Limit Theorem states that for large n, the distribution of the standardised sample mean Z =

(X )/(/ n) is approximately N(0, 1), irrespective of the distribution of the Xi .

As a rule of thumb, we consider n 30 to be large enough for the N(0, 1) distribution to provide
a close approximation. The approximation improves as n becomes larger.
5. The failure time of a particular type of electronic circuitboard is known to observe a Weibull
distribution with shape parameter = 2 and characteristic life = 5. We obtain a simple
random sample of fifty such circuitboards, and subject the boards to lifetime testing under identical
conditions. Denote the failure times by T1 , . . . , T50 .
(a) Find P(T1 > 5).
1 P50
(b) What is the probability distribution of the sample mean T = 50
i=1 Ti ? (Your answer
should state the values of the distribution parameters: use (1.5) = 0.8862 to obtain these.)

(c) Is this distribution exact, or an approximation? Give reasons.

(d) Find an approximate value for P(T > 5).
(e) Denote by the mean of the lifetime distribution. What is the minimum sample size needed
to ensure that P(| T |< 0.5) > 0.95?

(a) Since each Ti is Weibull(2, 5), the reliability function (for each component) is

R(t) = exp (t/)

= exp (t/5)2 ,
t 0.
P(T1 > 5) = R(5) = exp{12 } = exp{1} = 0.368 (3 d.p.).
(b) The mean and variance of the failure time distribution are

= 5 (1.5) = 4.431 (3 d.p.),
= 1 +


2 = 2 1 +


= 52 {(2) ((1.5))2 } = 25 {1! 0.88622 } = 5.366 (3 d.p.).

Since this is a sample random
P50 sample from a non-Normal population, the probability distribution of T = (1/50) i=1 Ti will not be exactly Normal. However, since n 30, the
distribution of T will be approximately N(, 2 /50) by the Central Limit Theorem. By substitution it follows that, approximately,

T =

1 X
Ti N(4.431, 0.107),

with the mean and variance correct to 3 decimal places.

(c) As we are sampling from a non-Normal population, the exact distribution of the sample mean
may be difficult to obtain. The Central Limit Theorem provides that the standardised sample
mean is approximately N(0, 1) in distribution for large sample sizes; this result is applied to
obtain the approximate distribution in (b).
(d) Using the approximate distribution for T obtained in (b),
P(T > 5) = P
/ 50
/ 50
P Z>
(where Z N(0, 1))
/ 50
5 4.431
= P Z>
= P(Z > 1.74) = 0.0409 (Tables).
(e) The requirement that

P | T | < 0.5 > 0.95
is (approximately) equivalent to

P |Z | <
/ n

> 0.95

(where Z N(0, 1)).

We see that the above probability is equal to 0.95 when


= z0.025 = 1.9600,
/ n
i.e. when n (1.96002 2 )/0.52 = 82.46 (2 d.p.). Thus, the requirement is satisfied for
samples of size n 83.