CVEN2002/CVEN2702

Engineering Computations - Statistics

Statistics Laboratory Class Week 1 : Introductory Matlab Tutorial
In your statistics course, there are scheduled computer tutorials in which you explore the use of
Matlab in statistics. Matlab is available on both the Linux and the Windows desktops in the
labs in the School of Mathematics and Statistics in the Red Centre (on the ground floor and on the
mezzanine level). Each student has a scheduled class with a tutor in the labs approximately every
second week for the session. Students are welcome to use the labs at any time when they are not busy
for other classes. Check notice boards for the current opening hours. You do need to be able to
use Matlab, so start as early as you can in the session!
Exercises and data sets for the computer tutorials will be made available on your course web page
(Moodle). Solutions to these exercises will usually be available from the course web page the following
week. There is a general Introduction to Matlab document available through the School of Mathematics and Statistics web page also available from your course web page. In addition there is an
Introduction to Matlab and Statistics Using Matlab (referred to here as SUM) document which is
available from the course web page. In all Matlab tutorials make sure you understand the
output you obtain. When you produce a graph, look at it! What does it tell you about the data?
Ask you tutor if you have questions. Explore the possibilities of Matlab - you may find it very useful
in all your courses!
Instructions
Pay particular attention to section 2 on plotting and graphing, and to section 3.1 where you use the Import
Wizard to input the data files for the computer tutorials in this course.

Double click on the Matlab icon and a Matlab Command Window should pop-up. The >> is a
prompt, requiring you to enter commands. One of the first commands you should type in is to find
out what your working directory is. The working directory is where you will save the results of your
calculations. So, enter
>> pwd
Note that pwd stands for print working directory. You probably would get an output like :
ans = /student/your last digit of your student id number/username
Matlab always stores the result of its last calculation in a variable called ans (short for answer).
Now this indicates the working directory.
Online help can be accessed for all Matlab commands by issuing the help command, or you can
use helpwin, helpdesk, or demo.
>> help htype name of command herei
To get started, you can simply type >> help
1. Variables, Vectors and Matrices
Variables in Matlab are like variables in any other programming language (C, C++ etc.), however,
one difference is that you do not have to define them by indicating the type. Also, variable names
are case sensitive and can be used to refer to a single number (a scalar) or an array of numbers (a
matrix or a vector). To create a row vector in Matlab, for example (note the spaces):
>> r = [1 2 3 4]
r =
1
2
3
4

A column vector can be created, for example, (note semicolons):

>> c = [1; 2; 3; 4]
c =
1
2
3
4
On the other hand, you can use the operator (transpose)
>> c = r
c =
1
2
3
4
Vectors can be created by incrementing a starting value with a constant quantity. For example (note
the use of colons):
>> r = [0:2:10]
r =
0
2
4
6
8
10
This creates a row vector, with the first element = 0 and each element incremented by 2; until the
final value of 10. You can index specific parts of a vector. For example, to get the third element in
the vector r:
>> r(3)
ans =
4
Matrices are 2 dimensional quantities and are created similarly. Eg:
>> a = [1 2 3; 4 5 6; 7 8 9; 10 11 12]
a =
1
2
3
4
5
6
7
8
9
10
11
12
Matrix a is a 4 3 matrix (4 rows and 3 columns). We can also use the incrementation principle, for
example:
>> b = [0:2:10; 1:2:11]
b =
0
2
4
6
8
10
1
3
5
7
9
11
Matrix b is a 2 6 matrix. Entries of the matrix, for instance the entry in the 2nd row, 5th column
can be accessed using the notation:
>> b(2, 5)
ans =
9
To extract data, for example the first column or second row of the matrix, use colons:
>> a(:,1)
ans =
1
4
7
10
(so this gives us the first column in the matrix a)

>> a(2,:)
ans =
4
5

(so this gives us the second row in the matrix a)

1.1. Vector and Matrix Operations. The basic arithmetic operations +, -, *, / can be used for
vectors and matrices. These would generate corresponding output vectors or matrices. For example,
to add two vectors:
>> a = [1 2 3 4];
>> b = [5 6 7 8];
>> c = a+b
c =
6 8 10 12
NOTE THAT the semicolons (;) in the first two commands direct Matlab not to echo the
values of the variables a and b on to the screen immediately after you type them. Check
that you understand what this means! What is the difference when you repeat the previous three
commands and dont put semicolons at the end of the commands? What if you do put a semicolon at
the end of the command c = a+b ?
As you know from algebra, only vectors (or matrices) that have the same size can be added or subtracted. For example:
>> a = [1:3:20; 21:3:40];
>> b = [2:3:20; 22:3:40];
>> c = a-b
c =
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
Matrix multiplication using the * symbol is possible only if the number of columns in the first matrix
equals the number of rows in the second:
>> a=[1 2 3; 4 5 6]
a =
1
2
3
4
5
6
>> b = a
b =
1
4
2
5
3
6
>> c=a*b
c =
14
32
32
77
If you want to multiply corresponding elements of two matrices (which have the same size), then you
do an array multiply using the .* symbol. For example
>> a = [1 2 3 4; 5 6 7 8]; b=[2 2 2 2; 3 3 3 3];
>> c=a .* b
c =
2
15

4
18

6
21

8
24

2. Plotting
The plot command is used for generating 1-D (functions of one variable) plots. Do >> help plot
for complete details. For example:
Lets make a graph of y = sin(x), for x on the interval x = 0 to x = 10.
>> x = [0:0.1:10]; (here .1 is the increment)
>> y = sin(x); (notice how the sin function operates on each element
of the entire row vector x, to generate another row vector y)
>> plot (x, y)
Surface or 2-D (functions of two variables) plots are generated using the surf command. To clear a
plot, type in >> clf (clear figure)
2.1. New figure, and subplot commands. To generate another plot window, do >> figure. If
you want to have several plots on the same figure in a grid then you can use the subplot command.
For example, to obtain a graph of y = sin(x), then on a second separate figure a graph of y = cos(x),
then on a third separate figure, graphs of both functions side by side:
>> x = [0:0.1:10];
>> y1 = sin(x);
>> y2 = cos(x);
>> plot (x, y1)
>> figure
>> plot(x,y2)
>> figure
>> subplot(1,2,1)
>> plot (x, y1)
>> title(graph of sin(x))
>> subplot(1,2,2)
>> plot(x,y2)
>> title(graph of cos(x))
In the subplot command the first two coordinates state the size of the grid (in this example, a 1 2
grid of graphs), and the third coordinate says where to put the next requested plot.
3. File input/output
In many instances, the simplest way to import data into Matlab is to use the Import Wizard. All
you need to do is click Import Data. In this course we only require you to be able to import data using
the Import Wizard and this is the same whether you are running Matlab on Linux or on Windows.
3.1. Data files for these tutorials. There are various data files which we will use in this course.
These are available from the course web page (Moodle). You should now use the following instructions
and use the Import Wizard (Home Import Data) to check that you can import the following files
without any problem.
Instructions: For each data file, locate the file on the course web page and then save it to your own
directory on the computer. Save the files as .txt files. Exactly how you do this may depend on
which browser you are using for the course web page. You may be able to simply save it to a file or
you may find it easier to open a text file and cut and paste the data and then save it to your own
directory. Experiment with the Import Wizard so that you understand how it works. Make sure you
save the data files in a location youll be able to easily access in the future, as we will make exhaustive
use of them in these laboratories.
inflows.txt This data file contains two columns of data, one called Year and one called Inflows.
Select this data file and click Import Data. Alternatively, you can right-click on the data
file and select Import Data. You should see that there are the two columns of data, with
names at the top, and 25 pairs of values. The columns are separated by a tab so check that

Tab is selected at the top. You can edit the variable names as Year and Inflow. This will
create vectors in Matlab called Year and Inflow. Note that the names are case sensitive.
Finally click the green tick Import Selection.
To list the data you can enter >> Year. (What happens if you forget the capital letter?) Then
enter >> Inflow. To check you have all the data you might check the length of the vectors
by >> length(Year). (You should get 25.) To see the two columns you could enter >> [Year
Inflow]. (What happens if you enter >> [year;Inflow]?)
kevlar90.txt Locate this file and then save it to your own directory. Note there is only one
column and there is no name at the top. Import it using the Import Wizard. How long is the
column? The vector itself is called kevlar90. Check the length of the vector. (There should
be 101 values).
shearstrength.txt. Locate, save and import this data file. There is one column of data. How
many values are there? After importing the values, check you can list them on the screen.
porevolume.txt This data set contains just one variable in one column. Locate it, save it
and import it using the Import Wizard.
fusiondat.txt Locate, save and import this data file. There are two columns of data values.
abs.txt The file abs.txt contains data on internal deadband impedances for front and rear
sensors from an antilock braking system (ABS). Each row of the dataset corresponds to the
same unit, with the first measurement being the front sensor impedance and the second the
rear sensor impedance (in kohms). There are two columns. Import the dataset as it is: the
whole data set matrix is called abs.
concrete.txt Locate, save and import this data file. This data file has three columns of data.
rain.txt The file rain.txt contains rainfall (first column) and runoff (second column) measurements at Pontelagoscuro on the Po river in northeast Italy, for the 31 years 1918 to 1948.
Locate, save and import the data.
milk.txt This data file has two columns, x = milk production (kg/day) and y = milk protein
(kg/day) for a sample of Holstein Friesian cows. Locate, save and import the data.
presdat2012.txt This contains data for the 44 US presidents and the 27 Australian Prime
ministers at the start of 2012. There are three columns, with the names (in order of leadership
of their country), and ages and then country.
3.2. Saving data (and other ways of data input). Alternatively, the save and load commands
are used for saving data to disk or loading data from disk, respectively. To save values in a matrix or
vector, called, for instance, y, do:
>> y = [1 2 3 4; 5 6 7 8];
>> save y.txt y -ascii
The -ascii option ensures that the data is saved in ASCII form, so that it can be read by other programs - Notepad, WordPad, Kwrite (in Linux) etc. Examine the file y.txt using one of these programs.
The file is generated in the working directory. An ASCII data file dat.txt can be loaded using >> load
dat.txt This provides a variable called dat in the Matlab workspace. All manner of vector/matrix
operations can be performed on dat, just like any other variable.
For learning all the load/save options use >> help save and >> help load.

4. Executable files (m-files)

Executable files in Matlab are generated by storing a list of Matlab commands in a file given the
extension .m. These files are called m-files (or scripts). To create an m-file, use the New Script option.
As an example, type in the following commands in the m-File Editor Window:
x = [0:0.1:10];
y = cos(x);
plot(x,y)
xlabel(x)
ylabel(y)
title(A plot of y=cos(x))
Save the file using the Save Option in the m-File Editor Window. Call it, say, cosineplot.m. Now,
to run this program in Matlab, move over to the Matlab Command Window and just type in >>
cosineplot .

End of Introductory Laboratory Class (Week 1).

CVEN2002/CVEN2702
Engineering Computations - Statistics
Statistics Laboratory Class Week 2 : Data and Histograms
Instructions
After this tutorial you should be able to: Access Matlab, and use the help facilities; enter and manipulate
vectors and matrices; transform data; enter data and display appropriately a histogram with appropriate titles
and labels; comment appropriately on a graphical display of data; (draw a stemplot of data (on paper, not
using Matlab); construct a histogram using the histogram command with the Matlab default bins or your
own choice of bins.
Browse the general Introduction to Matlab document and the Statistics Using Matlab (here referred to
as SUM) document to find extra information. Remind that you can also use the help command at any time
to get information about any Matlab command.
Do not just enter Matlab commands! Make sure you understand the output you obtain from the commands you use. When you produce a graph, look at it! What does it tell you about the data? Ask you tutor if
you have questions. Explore the possibilities of Matlab, go beyond what is actually required!
Solutions to these exercises will be available from the course web page later.

Exercise 1*
The Hardap Dam is the largest dam in Namibia, with a surface area of approximately 25 square
kilometres, a capacity of 320 million cubic metres, and a dam wall 862 metres long. Table 1 shows
annual maximum floodpeak inflows to the dam, in cubic metres per second, for years 1962/63 to
1986/87. (Data from Metcalfe, Andrew V. (1997). Statistics in Civil Engineering. London: Arnold ;
New York: Wiley).
Observation Y ear Inf low
Observation
1
1962 1864
14
2
1963 44
15
3
1964 146
16
4
1965 364
17
5
1966 911
18
6
1967 83
19
7
1968 477
20
8
1969 457
21
9
1970 782
22
10
1971 6100
23
11
1972 197
24
12
1973 3259
25
13
1974 554
Table 1. Annual maximum floodpeak inflows

Y ear
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986

Inf low
1506
1508
236
635
230
125
131
30
765
408
347
412

to Hardap dam

a) Recover the vectors Year and Inflow, imported from the inflows.txt data set into Matlab in the
first lab class.
b) Construct a frequency histogram of the inflows using the histogram command. Give the graph a
title, and label the x- and y-axes appropriately. (Reference SUM p.7-9 or use the help command).
Explore the histogram command so that you understand how to label appropriately.
c) Create a new data vector with values equal to the logarithms of the inflows. Construct a histogram
of the new data vector with appropriate labels and titles.

d) The coefficient of skewness is a measure of asymmetry of the data. It is defined as

Pn
(xi x)3
g = i=1 3
ns
where s is the sample standard deviation. It gives an indication of departures from symmetry: a
positive value of g may indicate a long right tail in the data with values above the median tending
to be further away from the median than values below the median, while negative skewness may
indicate a long left tail with values below the median tending to be more extreme.
Use Matlab to calculate the skewness of the two datasets (see SUM p.6 or use help command).
Compare the shapes of the histograms for the inflows and the log transformed inflows, and comment
on the degree and type of skewness.
e) Without using Matlab, construct a stem and leaf plot for the log-transformed inflows (round to
the tenths place). Note that this is a paper-and -pencil exercise, but you can use Matlab to help
order the data using the sort command.
f) Compare the histograms and stem and leaf plots for the log transformed data. What information
is there which is contained in one and not the other?
g) Write an m.file to carry out some or all of this exercise.
Exercise 2*
The table below shows the age at taking office for the 44 US presidents to date.
President
W ashington
J.Adams
Jef f erson
M adison
M onroe
J.Q.Adams
Jackson
V anBuren
W.H.Harrison
T yler
P olk
T aylor
F illmore
P ierce
Buchanan
Table 2. Age

Age
President Age
57
Lincoln
52
61
A.Johnson
56
57
Grant
46
57
Hayes
54
58
Garf ield
49
57
Arthur
51
61
Cleveland
47
54
B.Harrison 55
68
Cleveland
55
51
M cKinley
54
49
T.Roosevelt 42
64
T af t
51
50
W ilson
56
48
Harding
55
65
Coolidge
51
at taking office for the 44 US

President
Age
Hoover
54
F DRoosevelt 51
T ruman
60
Eisenhower
61
Kennedy
43
L.B.Johnson 55
N ixon
56
F ord
61
Carter
52
Reagan
69
G.W.H.Bush 64
Clinton
46
G.W.Bush
54
Obama
47
presidents to date

a) Recover the vectors name, age and country, imported from the presdat2012.txt data set into Matlab in the first lab class. Extract the ages of the US presidents only. Hint: the command strcmp
is used to compare character strings. The command A=age(strcmp(country,US)) will create a
vector A that contains age values for which country=US.
b) Construct a titled and labelled plot of president age versus time order. Comment on any temporal
patterns you see (i.e., what is the pattern, if any, over time).
c) Construct a titled and labelled frequency histogram of president ages.
d) Use the histogram command to construct an appropriately labelled frequency histogram showing
numbers of presidents whose ages fall in the ranges 42-44, 45-47, 48-50, ...,69-71. Hint: enter help
histogram for information on this command. You need to set up an edges vector to use in the
histogram function.

e) Compare the two frequency histograms produced in the previous steps. Comment briefly on the
visual impact of the choice of class intervals used in both.
f) Write an m.file to carry out some or all of this exercise.

End of Laboratory class for Week 2.

If you are unable to finish during the class then try to finish in your own time before your next lab
class so that you can seek help then if needed.

CVEN2002/CVEN2702
Engineering Computations - Statistics
Statistics Tutorial Class Week 3 : Descriptive Statistics
Instructions
The following are exercises that will covered during your tutorial class.
Note that not every topic in the course can be covered by the tutorial exercises. You are expected
to take an active part in the learning process by working by yourself on most of the topics. In particular, you
are expected to attempt these tutorial questions beforehand. During the tutorial, the tutor will go
through the answers to some of the questions while directing explanation to areas where students indicate they
have difficulty.
Solutions to these exercises will be available from the course web page at a later date.

Exercise 1*
a) Answer the following yes/no questions, and explain your answer.
(i) Will the sample mean always correspond to one of the observations of the sample?
(ii) Will exactly half of the observations in a sample always fall below the mean?
(iii) Will the sample mean always be the most frequently occurring data value in the sample?
(iv) Can the sample standard deviation be equal to zero?
(v) Can the sample median be equal to the sample mean?
b) (i) Suppose that you add 10 to all of the observations in a sample. How does this change the
sample mean? How does it change the sample standard deviation?
(ii) Suppose that you multiply all of the observations in a sample by 2. How does this change the
sample mean? How does it change the sample standard deviation?
(iii) A sample of temperature measurements in a furnace yielded a sample average of 446Celsius
and a sample standard deviation of 5.8Celsius. You would like to communicate this information to an American colleague. What are the sample average and the sample standard
deviation expressed in Fahrenheit? (Hint : temperature in C = (temperature in Fahrenheit
- 32)5/9)
Exercise 2*
An experiment to investigate the survival time (in hours) of an electronic component consists of
placing the parts in a test cell and running them for 100 hours under elevated temperature conditions
(this is called an accelerated life test). Eight components were tested with the following resulting
failure times :
75 63 100+ 36 51 45 80 90
The observation 100+ indicates that the unit still functioned at 100 hours. Is there any meaningful
measure of location that can be calculated for these data? What is its numerical value?
Exercise 3
Pn
1
2
Consider a sample of observations x1 , x2 , . . . , xn . For what value a is the quantity n1
i=1 (xi a)
minimised? Interpret in terms of the location and dispersion parameters that you know.
Exercise 4*
The following data is a sample of shear strength, (MPa) of a joint bonded in a particular manner :
22.4, 40.4, 16.4, 73.7, 36.6, 109.9, 30.0, 4.4, 33.1, 66.7, 81.5
a) Determine the 5-number summary.
b) Determine the iqr. Are there any outliers (by the 1.5 iqr rule)?

c) Construct a box-plot and comment on its features.

d) Determine the mean x and the sample standard deviation s.
e) By how much could the largest observation be decreased without affecting the iqr?
Exercise 5
a) The female students in an undergraduate engineering course self-reported their heights (in inches)
to the nearest inch. The data are :
62 64 66 67 65 68 61 65 67 65 64 63 67 68 64 66 68 69 65 67 62 66 68 67 66 65 69 65 70 65 67 68
65 63 64 67 67
Calculate the sample mean, the sample standard deviation and the sample median of height.
b) In the same class, the male students self-reported their heights as follows :
69 67 69 70 65 68 69 70 71 69 66 67 69 75 68 67 68 69 70 71 72 68 69 69 70 71 68 72 69 69 68 69
73 70 73 68 69 71 67 68 65 68 68 69 70 74 71 69 70 69
Calculate the sample mean, the sample standard deviation and the sample median of height.
c) Construct a comparative stem-and-leaf plot for both genders. Comment on any important features
that you notice in this display.
Exercise 6
Direct evidence of Newtons universal law of gravitation was provided from a renowned experiment
by Henry Cavendish (1731-1810). In the experiment, masses of objects were determined by weighting,
and measured force of attraction was used to calculate the density of earth. The values of the earths
density estimated by Cavendish, expressed as a multiple of the density of water (1 g/cm3 ), are:
5.50 5.30 5.47 5.10 5.29 5.65 5.55 5.61 5.75 5.63 5.27 5.44 5.57 5.36 4.88 5.86 5.34 5.39 5.34 5.53 5.29
4.07 5.85 5.46 5.42 5.79 5.62 5.58 5.26
(Source : Philosophical Transactions, 17 (1798), 469)
a)
b)
c)
d)

Find the sample mean, the sample standard deviation and the sample median of these data.
Determine the iqr. Are there any outliers (by the 1.5 iqr rule)?
Construct a box-plot and comment on its features.
Would you suggest the sample mean or the sample median as single estimate of the density of earth
from Cavendishs data?
Exercise 7

A.A. Michelson (1852-1931) made many series of measurements of the speed of light. Using a
revolving mirror techniques, he obtained
12 30 30 27 30 39 18 27 48 24 18
for the differences (velocity of light in air) - 299,700 km/s. (Source : The Astrophysical Journal, 65
(1927), 11.)
a) Draw a dotplot.
b) Find the median and the mean. Locate both on the dotplot.
c) Find the variance and standard deviation.
d) Find the quartiles.
e) Find the minimum, maximum, range, and interquartile range.
Exercise 8
An experimental study of the atomisation characteristics of biodiesel fuel was aimed at reducing the
pollution produced by diesel engines. Biodiesel fuel is recyclable and has low emission characteristics.
One aspect of the study is the droplet size (m) injected into the engine, at a fixed distance from the
nozzle. Consider the following observed droplet size :
2.1 2.2 2.2 2.3 2.3 2.4 2.5 2.5 2.5 2.8 2.9 2.9 2.9 3.0 3.1 3.1 3.2 3.3 3.3 3.3 3.4 3.5 3.6 3.6 3.6 3.7 3.7
4.0 4.2 4.5 4.9 5.1 5.2 5.3 5.7 6.0 6.1 7.1 7.8 7.9 8.9
(Source : Kim et al (2008), Energy and Fuels, 22, 20912098.)

a) Group these droplet sizes and obtain a frequency table using [2, 3), [3, 4) and [4, 5) as the first three
classes, but try larger classes for the other cases.
b) Construct a density histogram.
c) Obtain the sample mean x and the sample variance s2 .

End of the tutorial class for Week 3.

CVEN2002/CVEN2702
Engineering Computations - Statistics
Statistics Laboratory Class Week 4 : Descriptive Statistics
Instructions
After this tutorial you should be able to: draw density histograms; comment on density histograms; compute
the sample mean, the sample variance, the sample standard deviation and the five-number summary of a data
set; draw boxplots and side-by-side boxplots.
Browse the general Introduction to Matlab document and the Statistics Using Matlab (here referred to
as SUM) document to find extra information. Note that you can also use the help command at any time to
get information about any Matlab command.
Do not just enter Matlab commands! Make sure you understand what you do and the output you obtain.
When you produce a graph, look at it! What does it tell you about the data? Ask you tutor if you have
questions. Explore the possibilities of Matlab. Go beyond what is actually required!
Solutions to these exercises will be available from the course web page at a later date.

Exercise 1*
The data on inflows to the Hardap Dam (see Lab 2 Exercise 1) have been stored in the text file
inflows.txt. Recover the vectors Year and Inflow which were imported from the inflows.txt data set
into Matlab in the first lab class.
a) Use Matlab to find the mean, the variance, the standard deviation and the five-number summary
of the inflows.
b) Construct a density histogram to display the Inflows data. Give the graph a title, and label the
axes appropriately. The histogram function can construct a density histogram using the option
Normalization, pdf. For example to plot a density histogram for data x with 5 classes you
would type: histogram(x,5,Normalization,pdf)
c) How does the number of classes changes the histogram? Plot density histograms with 5 and 50
classes and comment on how informative they are.
d) Equal-width classes may not be a good choice if a data set stretches out to one side or the other
- as it is the case for the inflows data set. Add an edges vector to the histogram function to plot
a density histogram with classes [0, 500), [500, 1500), [1500, 3000) and [3000, 6500).
e) Comment on the shape of the histogram. Without looking at the summary statistics, which one
do you think is larger : the mean or the median? Why?
f) Construct a frequency histogram with the same classes as in c). Compare it to the density histogram
obtained in c). Which one is misleading and why?
g) Write an m.file to carry out some or all of this exercise.

Exercise 2
The file presdat2012.txt contains data for the 44 US presidents and 27 Australian prime ministers
to date giving the age at taking office. Recover the vectors Name, Age and Country, imported from
the presdat2012.txt data set into Matlab in the first lab class.
a) Calculate the sample mean, the sample variance and the sample standard deviation for ages at
taking office in the two countries and comment. Remember the command strcmp to compare
character string (see Lab 2, Exercise 2).
b) Construct side-by-side horizontal boxplots of age for the US and Australian leaders. Type help
boxplot if you are unsure how to do this.
c) Calculate the sample median, the sample quartiles and the sample interquartile range for each
country. Identify these on the boxplots.
d) The boxplot for the US shows two outlying values. Find the names of these two presidents who
were significantly older than the others when they took office.
e) Do the boxplots suggest that there is a difference in ages at taking office for leaders in the two
countries?
f) Write an m.file to carry out some or all of this exercise.

Exercise 3*
In a study designed to test the safety of various types of vehicles, vehicles were crashed into a wall
at 60 km/h with a crash-test dummy strapped in the drivers seat. In the crash.txt data file, you will
find the 6 variables
Head: measure of head injuries
Chest: chest deceleration (in g), which is a measure of chest injury
Airbag: 1 indicates an airbag, 0 indicates only seatbelts were used
Doors: number of doors (equals 6 for vans and sport utility vehicles)
Year: year of vehicle
Weight: weight of vehicle
Recover the vectors Head, Chest, Airbag, Doors, Year and Weight imported from the crash.txt data
set into Matlab from Moodle.
a)

i) Chest is a measure of chest injuries sustained by the crash-test dummy. Larger values indicate
more severe injuries. Use Matlab to find the mean, the variance, the standard deviation and
the five-number summary of the chest injuries.
ii) Determine the proportion of observations that have a value of Chest that is larger than or
equal to 60.
iii) Construct a density histogram for Chest. How many classes would you consider? Comment on
its shape. Experiment with changing the number of classes. For instance, produce a density
histogram with 6 classes, and one with 25 classes. Compare and comment.

b) Suppose we wanted to see if airbags prevented injuries. Since Airbag is a categorical variable and
Chest is a quantitative variable, a good form of analysis would be a side-by-side boxplot. Type
help boxplot if you are unsure how to do this.
i) Compare the distribution of chest deceleration for airbag (Airbag = 1) and non-airbag (Airbag
= 0) cars. Do airbags appear to prevent injury?
ii) How do the spreads of the two distributions compare?
iii) Are there any outliers?
iv) Is either distribution skewed? If so, describe the skewness.
c) Now lets compare the chest injuries of vehicles with different numbers of doors. Create side-by-side
boxplots, as above, using these two variables.
i) Does there appear to be a relationship between the number of doors and chest deceleration?
What about the plot suggests a relationship?
ii) Which type of vehicle (in terms of number of doors) tends to have the least severe injuries?
iii) Why do you think that chest deceleration might be related to number of doors on the vehicle?
d) To test one reasonable theory, make a boxplot of Weight by Doors. What do these boxplots suggest
about the results of the previous boxplot comparing chest deceleration by number of doors?
e) Write an m.file to carry out some or all of this exercise.

End of Laboratory class for Week 4.

If you are unable to finish during the class then try to finish in your own time before your next lab
class so that you can seek help then if needed.

CVEN2002/CVEN2702
Engineering Computations - Statistics
Statistics Tutorial Class Week 5 : Probabilities and Random Variables
Instructions
The following are exercises that will covered during your tutorial class.
Note that not every topic in the course can be covered by the tutorial exercises. You are expected
to take an active part in the learning process by working by yourself on most of the topics. In particular, you
are expected to attempt these tutorial questions beforehand. During the tutorial, the tutor will go
through the answers to some of the questions while directing explanation to areas where students indicate they
have difficulty.
Solutions to these exercises will be available from the course web page at a later date.

Exercise 1
For any collection of events A1 , A2 , A3 , . . . , Ak , it can be shown that the inequality
P(A1 A2 A3 A4 . . . Ak ) P(A1 ) + P(A2 ) + . . . + P(Ak )
always holds (try to understand why! Hint: use Venn diagrams and the additive law of probability).
This inequality is most useful in cases where the events involved have relatively small probabilities. For
example, suppose a system consists of five subcomponents connected in series and that each component
has a 0.01 probability of failing. Find the upper bound on the probability that the entire system fails.
For any collection of events A1 , A2 , A3 , . . . , Ak , it can be shown that the inequality


P(A1 A2 A3 A4 . . . Ak ) 1 P(Ac1 ) + P(Ac2 ) + . . . + P(Ack )
always holds (try to understand why! Hint: use De Morgans laws and the first part of this exercise).
This inequality is most useful in cases where the events involved have relatively high probabilities. For
example, suppose a system consists of ten subcomponents connected in series and that each component
has a 0.999 probability of functioning without failure. Find the lower bound on the probability that
the entire system functions correctly.
Exercise 2
Among the students doing a given course in engineering, there are four males enrolled in the civil
engineering program, six females enrolled in the civil engineering program, and six males enrolled in the
chemical engineering program. How many girls must be enrolled in the chemical engineering program
if gender and engineering program are to be independent when a student is selected at random?
Exercise 3*
You are imprisoned in a dungeon together with two fellow prisoners. You are informed by the jailer
that one of you has been chosen at random to be hanged, and the other two are to be freed. You ask
the jailer to tell you privately which of your fellow prisoners will be set free, claiming that there would
be no harm in divulging this information, since you already know that at least one of them will go
free.
a) The jailer refuses to answer the question, pointing out that if you knew which of your fellows were
to be set free, then your own probability of being executed would rise from 1/3 to 1/2, since you
would then be one of two prisoners. Show that the jailer is wrong.
b) You convince the jailer, and he tells you which of the other two will be set free. You say this
information to your fellow prisoners. While the spared guy is jumping for joy, the other one asks
you to switch your identities. Would you accept?

Exercise 4*
An oil exploration company currently has two active projects, one in Asia and the other in Europe.
Let A be the event that the Asian project is successful and B the event that the European project is
successful. Suppose that A and B are independent events with P(A) = 0.4 and P(B) = 0.7.
a) If the Asian project is not successful, what is the probability that the European project is also not
successful?
b) What is the probability that at least one of the two projects will be successful?
c) Given that at least one of the two projects is successful, what is the probability that only the Asian
project is successful?
Exercise 5
Let X be a r.v. with cumulative distribution function F (x) and density f (x) = F 0 (x). Find the
probability density function of
a) the maximum of n independent random variables all with cumulative distribution function F (x).
b) the minimum of n independent random variables all with cumulative distribution function F (x).
Exercise 6
An article in the review Knee Surgery, Sports Traumatology and Arthroscopy in 2005 cites a success
rate of more than 90% for meniscal tears with a rim width of less than 3mm, but only a 67% success
rate for tears of 36mm. If you are unlucky enough to suffer from a meniscal tear of less than 3mm
on your left knee and one of width 36mm on your right knee, what is the probability mass function
of the number of successful surgeries? Assume the surgeries are independent. Find the mean and
variance of the number of successful surgeries that you would undergo.
Exercise 7*
The article Error Distribution in Navigation (J. Institute of Navigation, 1971) suggests that the
distribution of the lateral position error, say X (in nautical miles), which can be either positive or
negative, is well approximated by a density like
f (x) = c e0.2 |x|

for < x < ,

for a constant c.
a) Find the value of c which makes f a legitimate density function, and sketch the corresponding
density curve.
b) In the long-run, what proportion of errors is negative? At most 2? Between 1 and 2?
Exercise 8
The probability density function of the weight X (in kg) of packages delivered by a post office is
70
for 1 < x < 70
f (x) =
69x2
and 0 elsewhere.
a) Determine the mean and the variance of the weight X.
b) If the shipping cost is $2.50 per kg, what is the average shipping cost of a package? What is the
variance of the shipping cost?
c) In the long-term, what is the proportion of packages whose weight exceeds 50 kg?
Exercise 9*
a) Given the four scatter plots for two random variables X and Y in Figure 1,
i) which of the plots demonstrates a positive relationship ?
ii) which of the plots demonstrates a positive linear relationship ?
iii) which of the plots demonstrates a negative relationship ?
iv) which of the plots demonstrates no relationship ?

Figure 1
v) for which of the plots would you expect positive correlation ? Negative correlation ? No or
little correlation ?
b) For each of the following pairs of variables, indicate whether you would expect a positive correlation,
a negative correlation, or little or no correlation. Explain your choices.
i) Maximum daily temperature and cooling cost
ii) Interest rate and number of loan applications
iii) Distance a student doing MATH2099 lives from UNSW campus and their marks at the Matlab
online quizzes
Exercise 10
Many families of probability distributions are useful for engineers, including Binomial and Poisson.
a) In each of the following situations state whether it is reasonable to use one of these distributions
for X. If so, tell which one, and (if it is possible) determine what are the values of the parameters :
i) Toss a fair coin 6 times, X is the number of Heads.
ii) Toss a fair coin until the first time a head appears, X is the count of the number of tosses you
make.
iii) A factory makes carpets. Sometimes there are flaws in the carpet. On average a square metre
of carpet has 3 flaws. X is the number of flaws in a random square metre of carpet.
iv) Most calls made at random by sample surveys dont succeed in talking to a person. Of calls in
New York City, only 1/12 succeed. A survey calls 500 randomly selected numbers in New York
City, and X is the number that reach a live person.
v) Calls to a telephone exchange come in at an average of 250 an hour, X is the number of calls
in a given hour.
vi) A die (6 faces, numbered 1,2,3,4,5,6) is tossed twice and X is the number of 6s obtained.
vii) A die (6 faces, numbered 1,2,3,4,5,6) is tossed 7 times and X is the largest number obtained
in the 7 tosses.
viii) A die (6 faces, numbered 1,2,3,4,5,6) is tossed 4 times and X is the total of the numbers
obtained on the top face.
b) Select one of the variables in part b) which can be modelled by a Binomial distribution, and
find the long-run proportion of times that X 1.

c) Select one of the variables in part b) which can be modelled by a Poisson distribution, and find
the long-run proportion of times that X 1.
d) In October 1994, a flaw in a certain Pentium chip installed in computers was discovered that could
result in a wrong answer when performing division. The manufacturer initially claimed that the
chance of any particular division being incorrect was only 1 in 9 billion, so that it would take
thousands of years before a typical user encountered a mistake. However, statisticians are not
typical users; some modern statistical techniques are so computationally intensive that a billion
divisions over a short period of time in not outside the realm of possibility. Assuming that the 1 in
9 billion figure is correct and that results of divisions are independent of one another, what is the
probability that at least one error occurs in one billion divisions with this chip?
Exercise 11
An individual claims to have extrasensory perception (ESP). As a test, a fair coin is tossed ten
times, and he is asked to predict in advance the outcome. Our individual gets seven out of ten correct.
What is the probability he would have done at least this well if he had no ESP? Would you believe in
his powers?
Exercise 12
a) Suppose a value Z is repeatedly randomly chosen from a standard normal distribution:
i) In the long run, what is the proportion of times that Z will be at most 2.15? Less than 2.15?
ii) What is the long run proportion of times that Z will be between -1.23 and 2.85?
iii) What is the long run proportion of times that Z will exceed 5? Will exceed -5?
iv) What is the long run proportion of times that Z will satisfy |Z| < 2.50?
b) The article Characterization of Room Temperature Damping in Aluminum-Indium Alloys (Metallurgical Trans., 1993) suggests that Al matrix grain size (m) for an alloy consisting of 2% Indium
could be modeled with a normal distribution with a mean value 96 and standard deviation 14.
i) What is the probability that grain size exceeds 100?
ii) What is the probability that grain size is between 50 and 80?
iii) What interval (a, b) includes the central 90% of all grain sizes (so that 5% are below a and 5%
are above b)?

End of the tutorial class for Week 5.

CVEN2002/CVEN2702
Engineering Computations - Statistics
Statistics Laboratory Class Week 6 : Probability and Distributions
Instructions
After this tutorial you should be able to: use Matlab for determining probabilities from common probability
distributions.
Browse the general Introduction to Matlab document and the Statistics Using Matlab (here referred to
as SUM) document to find extra information. Remind that you can also use the help command at any time
to get information about any Matlab command.
Do not just enter Matlab commands! Make sure you understand what you do and the output you obtain.
When you produce a graph, look at it! What does it tell you about the data? Ask you tutor if you have
questions. Explore the possibilities of Matlab, go beyond what is actually required!
Solutions to these exercises will be available from the course web page at a later date.

Matlab has pdf and cdf commands for determining probabilities from many distributions, including Normal, Exponential, Uniform, Binomial and Poisson. Use the help command for more
details, or see Section 4 of the SUM. Note that Matlab uses the notation pdf for both the probability mass function (pmf) of discrete distributions and the probability density function (pdf) of
continuous distributions.

Exercise 1: Probability of events

Consider tossing a fair die. Define the events A = {2, 4, 6} and B = {1, 2, 3, 4}. Then, P(A) = 1/2,
P(B) = 2/3 and P(A B) = 1/3 (check this!). This shows that the events A and B are independent
(why?). Using the Matlab command randi (use help to determine how to use it), simulate draws
from the sample space a large number of times (say, N = 1, 000 times) and estimate the probabilities
by the observed proportions pA , pB and pAB . Verify that pA pB ' pAB . Now find two events which
are not independent. Repeat.
Exercise 2: the Binomial distribution*
The Binomial distribution is a discrete distribution. Use help to determine how to use binopdf
and binocdf.
a) For Y Bin(20, 0.35), compute
i) P(Y = 5)
ii) P(Y = 0)
iii) P(Y 0), and compare ii)
iv) P(5 Y < 15)
v) P(5 < Y 15)
vi) P(Y {1, 5, 9, 17})
vii) P(Y
/ {5, 13, 16})
b) The multiple-choice Statistics Online Quiz 1 contained 10 questions, each with 5 answers. Assume
a student just guessed on each question.
i) What was the probability that the student answered all the questions correctly?
ii) What was the probability that the student answered at least 5 questions correctly?
Suppose the three Statistics Online Quizzes over the semester are similarly designed (10 questions,
each with 5 answers)
iii) What is the probability that the student, guessing on each question in each quiz, passes (at
least 5 correct answers out of 10 questions) on at least one quiz?

Exercise 3: the Poisson distribution*

The Poisson distribution is a discrete distribution. Use help to determine how to use poisspdf and
poisscdf.
a) For Q P(7), compute
i) P(Q 3)
ii) P(Q < 3)
iii) P((Q = 19) (Q = 20))
iv) P(Q > 0)
b) Astronomers treat the number of stars in a given volume of space as a Poisson random variable.
On average, in the Milky Way galaxy in the vicinity of our Solar System, they find one star per 16
cubic light-years.
i) What is the probability of two or more stars in 16 cubic light-years?
ii) How many cubic light-years of space must be studied so that the probability of one or more
stars exceeds 0.95? (Hint: recall the Poisson process (Slide 29 and 33, Lecture 5))
Exercise 4: the Uniform distribution*
The Uniform distribution is a continuous distribution. Use help to determine how to use unifpdf
and unifcdf. You may also find useful the command unifinv (use help to find out what it is for).
a) For X U[1,1] , find
i) P(X < 0)
ii) P(X 0)
iii) P(0.9 X 0.8)
iv) the value of x such that P(x X x) = 0.9
b) The thickness of photoresist applied to wafers in semiconductor manufacturing at a particular
location on the wafer is uniformly distributed between 0.2050 and 0.2150.
i) Determine the proportion of wafers that exceeds 0.2125 micrometers in photoresist thickness.
ii) What thickness is exceeded by 10% of the wafers?
Exercise 5: the Exponential distribution*
The Exponential distribution is a continuous distribution. Use help to determine how to use exppdf
and expcdf. You may also find useful the command expinv (use help to find out what it is for).
a) For
i)
ii)
iii)
iv)

W Exp(2), find
P(W 2)
P(W < 2)
P(10 < W < 13)
P(W > 5)

b) Suppose that the time to failure (in hours) of fans in a personal computer can be modeled by an
exponential distribution with = 1/0.0003.
i) What proportion of fans will last at least 10,000 hours?
ii) What proportion of fans will last at most 7000 hours?
iii) 95% of the fans will last for at least which amount of time?
Exercise 6: the Normal distribution*
The Normal distribution is a continuous distribution. Use help to determine how to use normpdf
and normcdf. You may also find useful the command norminv (use help to find out what it is for).
a) For Z N (0, 1), compute
i) P(1 < Z < 1)
ii) P(2 < Z < 2)
iii) P(3 < Z < 3)
iv) Comment on i), ii) and iii) (look at Slide 44 Lecture 5)
v) What is the value of z such that P(Z > z) = 0.05?

vi) What is the value of z such that P(Z > z) = 0.025?

vii) What is the value of z such that P(Z > z) = 0.005?
viii) Comment on v), vi) and vii) (look at Slide 51 Lecture 5)
b) For
i)
ii)
iii)
iv)

X N (3, 2), compute

P(2 < X < 4)
P(2 X 4)
P(X 4)
P(1 < X < 5), and compare to a) i).

c) An article in the review Archives of Environmental and Occupational Health considered polycyclic
aromatic hydrocarbons and immune system function in beef cattle. Some cattle were near major oiland gas-producing areas of western Canada. The mean monthly exposure to PM1.0 (particulate
matter that is 1m in diameter) was approximately 7.1 g/m3 with standard deviation 1.5
g/m3 . Assume the monthly exposure is normally distributed.
i) What is the probability of a monthly exposure greater than 9 g/m3 ?
ii) What is the probability of a monthly exposure between 3 and 8 g/m3 ?
iii) What is the monthly exposure that is exceeded with probability 0.05?
Exercise 7: the disttool command in Matlab
a) Start the Matlab demo on distributions by typing disttool in the Command Window. A graph
showing the cumulative distribution function (CDF) for the standard normal distribution should
appear. You can view the probability density function (PDF) by changing the Function type from
CDF to PDF. To change the mean and standard deviation you can use the sliding bars labelled Mu
() and Sigma (), or type in values. Experiment with these features to see the effect of changing
the mean and standard deviation.
b) What happens to the pdf as the mean is increased or decreased?
c) What happens to the pdf as the standard deviation is increased or decreased?
d) For the normal random variable N (3, 1), find the value of the cdf at x = 2. (Hint: change the
mean and standard deviation, then type the value 2 in the X box, click anywhere in the grey area
outside the graph to update it, and look at the value in the Probability box). Interpret this value
in terms of a probability statement.
e) For the normal random variable N (3, 1) find the value of x that cuts off probability 0.05 in the
upper tail (Hint: this will be 0.95 in the lower tail. Use the Probability box). Interpret this value
in terms of a probability statement.
f) Experiment with the disttool tool with the other distributions introduced in class: Binomial,
Poisson, Uniform and Exponential.

End of Laboratory class for Week 6.

If you are unable to finish during the class then try to finish in your own time before your next lab
class so that you can seek help then if needed.

CVEN2002/CVEN2702
Engineering Computations - Statistics
Statistics Laboratory Class Week 8 :
The Central Limit Theorem and Confidence Intervals
Instructions
After this tutorial you should be able to: use simulations in Matlab to explore the sampling distribution
of the sample mean and to illustrate the Central Limit Theorem, use Matlab to derive confidence intervals
of a given level, and check if the normal assumption is reasonable.
Browse the general Introduction to Matlab document and the Statistics Using Matlab (here referred to
as SUM) document to find extra information. Remind that you can also use the help command at any time
to get information about any Matlab command.
Do not just enter Matlab commands! Make sure you understand what you do and the output you obtain.
When you produce a graph, look at it! What does it tell you about the data? Ask your tutor if you have
questions. Explore the possibilities of Matlab. Go beyond what is actually required!
Solutions to these exercises will be available from the course web page at a later date.

In Matlab you can simulate random samples from known distributions, using a -rnd command (rnd
for random). For instance, you can obtain a sample column vector of 5 values from the Normal distribution N (12, 3) by entering the command normrnd(12,3,5,1). If you just enter normrnd(12,3,5),
then you will get a 5 5 matrix of random values from N (12, 3). To obtain samples from the Exponential distribution, use exprnd, to obtain samples from the Uniform distribution, use unifrnd,
to obtain samples from the Binomial distribution, use binornd, to obtain samples from the Poisson
distribution, use poissrnd. Use help for more information. See also Section 5 of the SUM document.
Exercise 1*
In this question we consider lots of simulated random samples of size 100 from a Binomial distribution.
a) Simulate (use the Matlab command binornd) a matrix B with 100 rows and 500 columns containing elements that are independent values drawn from the Binomial distribution with n = 1 and
= 0.5 (that is, the Bernoulli distribution with parameter = 0.5). You can think of each column
of B as being a random sample of size 100 from X Bern(0.5). With 500 columns in B, we have
500 independent random samples of size 100.
b) Use the Matlab mean function to compute the means of each of the 500 columns. You get a row
vector, call it meanB (say), of length 500, which consists of the means of 500 samples of size 100,
= 1 (X1 + X2 + . . . + X100 ).
that is 500 values of the random variable X
100
Note that, when applied to a matrix, the Matlab function mean computes the mean of each column
and returns a vector.
c) From lectures, what do you expect the mean and variance of the sampling distribution of the
to be? Calculate the sample mean and variance for meanB and compare with
random variable X
your expectations.
to be?
d) From lectures, what do you expect the sampling distribution of the random variable X
Recall the Central Limit Theorem. Construct a density histogram of the simulated sample means.
Comment on the shape.

e) The histfit function can overlay a histogram with the probability density function of any distribution. To plot a histogram of meanB with 10 classes, overlaid with the Normal distribution type
histfit(meanB,10,normal), see help histfit for other distributions. Comment on how well
the Normal distribution fits the data.
f) Write an m.file to carry out some or all of this exercise.
Exercise 2
The file kevlar90.txt contains times to failure (in hours) for a sample of size n = 101 of strands of
Kevlar 49/epoxy, a material used in the space shuttle, when tested at a stress level of 90%. Recover
the vector kevlar90, imported from the kevlar90.txt data set into Matlab in the first lab class.
a)

i) Construct a density histogram for the kevlar90 data. Comment on the shape. Do you think that
a Normal distribution would fit the data well? Do you think that an Exponential distribution
would fit the data well?
ii) Find the sample mean x of the 101 failure times.
iii) From i), we can assume that the Exponential distribution is the population distribution, and
that the sample is a random sample drawn from it. Recall that the expectation of the
Exponential()-distribution is , which can thus be regarded as the population mean. We
can estimate the true Exponential distribution Exp() by the Exponential distribution with
parameter x: we just replace in the distribution the unknown value of the mean () by its
natural estimate, the sample mean x (
= x). This simple way of fitting a distribution is
known as the method of moments. Write down the fitted Exponential density for the Kevlar
failure times. Given that the standard deviation of the Exp() density is also , state the
standard deviation of a random variable with the fitted Exponential density. Compare the
sample standard deviation of the data set kevlar90.
iv) Use the histfit function to overlay the histogram of the Kevlar data with the fitted Exponential density and briefly comment on how well the exponential density fits the data.
v) Write an m.file to carry out some or all of this exercise.

b) Next we examine the sampling distribution of the mean failure time for a sample of 10 strands of
Kevlar.
i) Use the Matlab command exprnd to simulate a matrix T with 10 rows and 1000 columns
containing independent elements that are Exponentially distributed with mean equal to the
sample mean of the Kevlar failure times. We can think of each column of T as being failure
times for a sample of 10 strands of Kevlar similar to those in the dataset. With 1000 columns
in T, we have 1000 random samples of size 10. Each of them could have been the observed
sample in kevlar90.txt! It must be clear that here the population distribution which generated
the samples is Exp(
), with
found in a)iii).
ii) Use the mean function to compute the means of the 1000 columns. Thus, you get a row vector,
called meanT (say), of length 1000, which consists of the means of 1000 samples of size 10, that
= 1 (X1 + X2 + . . . + X10 ).
is 1000 values of the random variable X
10
iii) From lectures, what do you expect the mean and variance of the sampling distribution of the
to be? Calculate the sample mean and variance for meanT and compare
random variable X
with your expectations.
to be?
iv) From lectures, what do you expect the sampling distribution of the random variable X
Recall the Central Limit Theorem. Construct a density histogram of the simulated sample
means. Comment on the shape.
v) Use the histfit function to overlay the expected large sample density on the histogram and
briefly comment on the level of similarity.
vi) Write an m.file to carry out some or all of this exercise.
c) Repeat part b) but this time generating 1000 random samples of size n = 500.
Exercise 3*
The file shearstrength.txt contains 100 observations on shear strength (in lb) of ultrasonic spot
welds made on a certain type of alclad sheet. This data is in agreement with the one in Comparison of
Properties of Joints Prepared by Ultrasonic Welding of Other Means, Journal of Aircraft, 1983: 552556. Recover the vector shearstrength, imported from the shearstrength.txt data set into Matlab
in the first lab class.
a) Construct a density histogram of the shear strengths. Comment on the shape of the distribution.
b) Is it plausible that the sample was selected from a normal distribution? Draw a normal quantile plot
and conclude. Note: in Matlab, a normal quantile plot is represented by entering the command
qqplot. Use help qqplot for more information.
c) Investigators believe that the population standard deviation of shear strength in this context is
= 350 lb. Estimate , the true mean shear strength and give a 95% two-sided confidence interval
for it. Interpret your result. Note: in Matlab, this type of z-confidence interval (see Slide 55,
Lecture 7 of the lecture slides) is obtained with the command ztest. This is actually a command
used for hypothesis testing (which we have not yet studied), but is also used for determining

confidence intervals. For instance, to calculate the 95% CI for a population mean from a sample
contained in the vector data, use the command
[h, p, ci] = ztest(data, xbar, sigma); ci
where xbar is the sample mean and sigma is the supposedly known population standard deviation
. The default option is for a 95% confidence interval. If you require any other level of significance
you need to enter a fourth parameter alpha. The outputs h and p refer to the underlying hypothesis
test and are not relevant for us at this point; that is why we focus only on the output ci here.
Type help ztest for more information. See also the SUM document, Section 6.1.
d) Now, suppose is not known. Give a 95% two-sided confidence interval for . Interpret your
result. Compare with the previous confidence interval, and explain why they are different. Note:
in Matlab, this type of t-confidence interval (see Slide 55, Lecture 7 of the lecture slides) is
obtained with the command ttest. Again, this is actually a command used for hypothesis testing
but it is also used for determining confidence intervals. For instance, to calculate the 95% CI for a
population mean from a sample contained in the vector data, use the command
[h, p, ci] = ttest(data, xbar); ci
where xbar is the sample mean. The default option is for a 95% confidence interval. If you require
any other level of significance you need to enter a third parameter alpha. Type help ttest for
more information. See also the SUM document, Section 6.1.
e) Write an m.file to carry out some or all of this exercise.
Exercise 4
In this exercise, we will use simulations to drive home the proper interpretation of confidence intervals.
a) Simulate (use the Matlab command normrnd) a matrix C with 36 rows and 500 columns containing
elements that are independent values drawn from the Normal distribution with = 20 and = 5.
You can think of each column of C as being a random sample of size 36 from X N (20, 5). With
500 columns in C, we have 500 independent random samples of size 36.
b) Calculate 95% z- and t-confidence intervals for from each sample. Recall that the z-confidence
interval of level 100 (1 )% is given by



x z1/2
n
where is assumed to be known, and the t-confidence interval of level 100 (1 )% is


s
x tn1;1/2
n
where s is the sample standard deviation.
i) Use the Matlab mean function to compute the means of each of the 500 samples. You get a
row vector, call it meanC (say), of length 500, which consists of the 500 sample means.
Note that, when applied to a matrix, the Matlab function mean computes the mean of each
column and returns a vector.
ii) Use the Matlab std function to compute the standard deviations of each of the 500 samples.
You get a row vector, call it sC (say), of length 500, which consists of the 500 sample standard
deviations.
(Like mean, when applied to a matrix, std computes the standard deviation of each column
and returns a vector.)

iii) From the vector meanC, build the vector upz of upper bounds for the z-CI for each sample.
Recall that z0.975 = 1.96 and that here, n = 36 and = 5 (if assumed to be known).
iv) From the vector meanC, build the vector lowz of lower bounds for the z-CI for each sample.
v) From the vectors meanC and sC, build the vector upt of upper bounds for the t-CI for each
sample. Find the value t35;0.975 in Matlab (hint: use the tinv function).
vi) From the vectors meanC and sC, build the vector lowt of lower bounds for the t-CI for each
sample.
c) You have now 500 z-confidence intervals and 500 t-confidence intervals for computed from 500
independent samples. Compute how many of the 500 z-confidence intervals and how many of the
500 t-confidence intervals contain the true population mean = 20. Comment.
d) If you were to repeat this exercise over and over, on average, what fraction of the confidence intervals
you calculate would you expect to contain the population mean?
e) Write an m.file to carry out some or all of this exercise.
Exercise 5
For each of 18 preserved cores from oil-west carbonate reservoirs, the amount of residual gas saturation after a solvent injection was measured at water flood-out. The data can be found in the
file porevolume.txt, and are from Relative Permeability Studies of Gas Water Flow Following Solvent Injection in Carbonate Rocks, Soc. Petroleum Engineers J., 1976: 23-30. Recover the vector
porevolume, imported from the porevolume.txt data set into Matlab in the first lab class.
a) Is it plausible that the sample was selected from a normal population distribution? Show a density
histogram and a normal quantile plot and conclude.
b) Determine a 98% CI for the true average amount of residual gas saturation.
c) Determine a 95% CI for the true average amount of residual gas saturation. What is the difference
between a 95% CI and a 98% CI?
d) Write an m.file to carry out some or all of this exercise.
Exercise 6
The file fusiondat.txt contains results from an experiment in visual perception using random dot
sterograms. Two images appear to be composed entirely of random dots, but the images will fuse
and a 3D object will appear when they are viewed in a certain way. An experiment was performed to
determine whether knowledge of the form of the embedded image affected the time required for subjects
to fuse the images. One group of subjects received either no information or just verbal information
about the shape of the embedded object. A second group received both verbal information and
visual information (e.g., a drawing of the object). Recover the matrix fusiondat, imported from the
fusiondat.txt data set into Matlab in the first lab class. Notice the form of the data file with the two
columns. The variables in the file are the group the subject belongs to (1 = no information or just
verbal information, 2 = verbal information and visual information), and the log of the time taken to
fuse the images.

a) Which formula do you think is appropriate for constructing a confidence interval for the difference
in mean log times for the two groups? Why?
b) Use this formula to construct a 95% confidence interval for the difference in mean log times for the
two groups. Does the provision of visual information appear to make a difference in time taken to
fuse the images?
c) Write an m.file to carry out some or all of this exercise.
Exercise 7
The file abs.txt contains data on internal deadband impedances for front and rear sensors from an
antilock braking system (ABS). Each row of the dataset corresponds to the same unit, with the first
measurement being the front sensor impedance and the second the rear sensor impedance (in kohms).
Recover the matrix abs, imported from the abs.txt data set into Matlab in the first lab class.
a) Which formula do you think is appropriate for constructing a confidence interval for the difference
in mean impedance for the front and rear sensors? Why?
b) Construct a 95% confidence interval for the difference in mean impedance for front and rear sensors.
Does location of sensor appear to affect impedance?

End of Computer Tutorial for Week 8.

If you are unable to finish during the class then try to finish in your own time before the lab test so
that you can seek help before it if needed.

CVEN2002/CVEN2702
Engineering Computations - Statistics
Statistics Tutorial Class Week 9 :
Estimation, Central Limit Theorem and Inferences concerning a mean
Instructions
The exercises that will be gone over during the tutorial class will be chosen from those printed here.
Note that not every topic in the course can be covered by the tutorial exercises. You are expected
to take an active part in the learning process, by working by yourself on most of the topics.
In particular, you are expected to attempt these tutorial questions beforehand. At the tutorial,
the tutor will go through the answers to some of the questions, directing explanation to areas where students
indicate they have difficulty.
Solutions to these exercises will be available from the course web page at a later date.

Exercise 1*
a) Suppose that X is a random variable which is Uniform on the interval [2, 5].
i) Write down the density function of X. What are , 2 and for this distribution?
ii) Random samples of size n = 15 are taken from the distribution of X. Determine the mean,

variance and standard deviation of the sampling distribution of the sample mean X.
b) Suppose that X is a random variable which is Exponentially distributed with mean 2.
i) Write down the density function of X. What are , 2 and for this distribution?
ii) Random samples of size n = 12 are taken from the distribution of X. Determine the mean,

variance and standard deviation of the sampling distribution of the sample mean X.
c) Suppose that X is a random variable which is Poisson distributed with mean 5.
i) Write down the probability mass function of X. What are , 2 and for this distribution?
ii) Random samples of size n = 10 are taken from the distribution of X. Determine the mean,

variance and standard deviation of the sampling distribution of the sample mean X.
Exercise 2
a) Suppose we have a random sample X1 , X2 , . . . , X2n of size 2n from a population denoted by X, and
E(X) = and Var(X) = 2 . Let
2n

X
1 = 1
X
Xi
2n i=1

and

X
2 = 1
X
Xi
n i=1

be two estimators of . Which is the better estimator of ? Explain.

b) Let X1 , X2 , . . . , Xn denote a random sample from a population having mean and variance 2 .
Consider the following estimators of :

1 =

X1 + X2 + . . . + X7
7

2 =

2X1 X6 + X4
2

i) Is either estimator unbiased?

ii) Which estimator is best? In what sense is it best?

c) Two different plasma etchers in a semiconductor factory have the same mean etch rate . However,
machine 1 is newer than machine 2 and consequently has smaller variability in etch rate. We
know that the variance of etch rate for machine 1 is 12 and for machine 2 is 22 = a12 (a > 1).
Suppose that we have n1 independent observations on etch rate from machine 1 and n2 independent
observations on etch rate from machine 2.
1 + (1 )X
2 is an unbiased estimator for for any value of between 0
i) Show that
= X
and 1.
ii) Find the standard error of the point estimate of in part i).
iii) What value of would minimise the standard error of the point estimate of ?
iv) Suppose that a = 4 and n1 = 2n2 . What value of would you select to minimise the standard
error of the point estimate of ? How bad would it be to arbitrarily choose = 0.5 in this
case? How bad would it be to only use the observations coming from machine 1?
Exercise 3*
a) A synthetic fibre used in manufacturing carpet has tensile strength that is normally distributed
with mean 75.5 psi and standard deviation 3.5 psi. Find the probability that a random sample of
n = 6 fibre specimens will have a sample mean tensile strength that exceeds 75.75 psi. How does
this probability change when the sample size is increased from n = 6 to n = 49?
b) The compressive strength of concrete is normally distributed with = 2500 psi and = 50 psi.
Find the probability that a random sample of n = 5 specimens will have a sample mean strength
that falls in the interval from 2499 to 2510 psi.
c) The amount of time that a customer spends waiting at an airport check-in counter is a random
variable with mean 8.2 minutes and standard deviation 1.5 minutes. Suppose that a random
sample of n = 49 customers is observed. Find the probability that the average waiting time for
these customers is
i) less than 10 minutes.
ii) between 7 and 10 minutes.
iii) less than 8.5 minutes.
Exercise 4
The Rockwell hardness of certain metal pins is known to have a mean of 50 and a standard deviation
of 1.5.
a) Suppose that it is known that the distribution of all such pin hardness measurements is normal,
then what can be said about the sampling distribution of the sample means from random samples
of size n? Under this assumption of normality, what is the probability that the average hardness
for a random sample of size n = 9 is at least 52?
b) What is the probability that the average hardness in a random sample of 40 pins is at least 52? Do
you need to assume normality here?
Exercise 5*
The number of flaws X on an electroplated car grill is known to have the following probability mass
function:
x
0
1
2
3
p(x) 0.8 0.1 0.05 0.05
a) Calculate the mean and standard deviation of X.
b) What are the mean and the standard deviation of the sampling distribution of the average number
of flaws per grill in a random sample of 64 grills?
c) For a random sample of 64 grills, calculate (approximately) the probability that the average number
of flaws per grill exceeds 0.5.

Exercise 6
An article in Journal of Agricultural Science (1997) investigated means of wheat grain crude protein
content (CP) and Hagberg falling number (HFN) surveyed in the UK. The analysis used a variety of
nitrogen fertiliser applications (kg N/ha), temperature (C), and total monthly rainfall (mm). The data
shown below describe temperatures for wheat grown at Harper Adams Agricultural College between
1982 and 1993. The temperatures measured in June were obtained as follows:
15.2 14.2 14.0 12.2 14.4 12.5 14.3 14.2 13.5 11.8 15.2
Assume that the standard deviation is historically known to be = 0.5 and that the temperature
distribution is normal.
a) Construct a 99% two-sided confidence interval on the mean temperature.
b) Construct a 95% lower-confidence bound on the mean temperature (that is, a one-sided confidence
interval).
c) Suppose that we wanted to be 95% confident that the error in estimating the mean temperature is
less than 0.2C. What sample size should be used?
Exercise 7
The wall thickness of 25 glass 2-litre bottles was measured by a quality-control engineer. The sample
mean was x = 4.05 millimetres, and the sample standard deviation was s = 0.08 millimetre. Find a
95% lower confidence bound for mean wall thickness. Interpret the interval you have obtained. Assume
the normal distribution for wall thickness.
Exercise 8
While performing a certain task under simulated weightlessness, the pulse rate of 42 astronaut
trainees increased by an average of 26.4 beats per minute with a standard deviation of 4.28 beats per
minute.
a) Construct a two-sided 95% confidence interval for the true average increase in the pulse rate of
astronaut trainees performing the given task.
b) What can one assert with 95% confidence about the maximum error if x = 26.4 is used as a point
estimate of the true average increase in the pulse rate?
Exercise 9*
Consider the following sample of fat content (in percentage) of n = 10 randomly selected hot dog
sausages of a given brand (Sensory and Mechanical Assessment of the Quality of Frankfurters,
Journal of Texture Studies, 1990):
25.2 21.3 22.8 17.0 29.8 21.0 25.5 16.0 20.9 19.5
Assume that the fat content follows a normal population.
a) Find a 95% confidence interval for the true mean fat content of the sausages of that brand.
b) Suppose, however, you are going to eat a single hot dog of this type and want a prediction for the
resulting fat content. Find a 95% prediction interval for the fat content of the hot dog sausage you
will eat.
Exercise 10
The article Repeatability and Reproducibility for Pass/Fail data (J. of Testing and Eval., 1997,
151-153) reported that in n = 48 trials in a particular laboratory, 16 resulted in ignition of a particular
type of substrate by a lighted cigarette. Find a 95% approximate confidence interval for , the true
long-run proportion (or probability) of all such trials that would result in ignition.

End of the tutorial class for Week 9.

CVEN2002/CVEN2702
Engineering Computations - Statistics
Statistics Tutorial Class Week 11 : Confidence Intervals and Hypotheses Tests
Instructions
The exercises that will be covered during the tutorial class will be chosen from those printed here.
Note that not every topic in the course can be covered by the tutorial exercises. You are expected
to take an active part in the learning process, by working by yourself on most of the topics.
In particular, you are expected to attempt these tutorial questions beforehand. During the tutorial,
the tutor will go through the answers to some of the questions, directing explanation to areas where students
indicate they have difficulty.
Solutions to these exercises will be available from the course web page at a later date.

Exercise 1*
a) In a hypothesis test of H0 with alternative Ha , the p-value is calculated to be 0.008. What would
be the conclusion of the hypothesis test?
b) In a hypothesis test of H0 with alternative Ha , the p-value is calculated to be 0.08. What would
be the conclusion of the hypothesis test?
c) In a hypothesis test of H0 with alternative Ha , the p-value is calculated to be 0.8. What would be
the conclusion of the hypothesis test?
Exercise 2*
In the book Statistical quality design and control: Contemporary concepts and methods DeVore,
Chang, and Sutherland (1992) discuss a cyclinder boring process for an engine block. Specifications
require that these bores be 3.5199 0.0004 inches. Management is concerned that the true proportion
of cylinder bores outside the specifications is excessive. Current practice is willing to tolerate up to
10% outside the specifications. Out of a random sample of 165 observations, 36 were outside the
specifications.
a) Conduct the most appropriate hypothesis test using a 0.01 significance level.
b) Construct a 99% confidence interval for the true proportion of bores outside the specification.
c) What assumptions did you need to make to carry out this hypothesis test, and determine the
confidence interval? Can you check these assumptions? If so, how?
Exercise 3*
The paper Selection of a Method to Determine Residual Chlorine in Sewage Effluents (Water and
Sewage Works, 1971, pp. 360-364) reports the results of an experiment in which two different methods
for determining chlorine content were used on specimens of Cl2 -demand-free water for various doses
and contact times. Observations are in mg/l.
Specimen
1
2
3
4
5
6
7
8
MSI method 0.39 0.84 1.76 3.35 4.69 7.70 10.5 10.92
SIB method 0.36 1.35 2.56 3.92 5.35 8.33 10.70 10.91
The sample mean of the differences (MSI-SIB) between readings for these 8 samples is 0.4137 and
the corresponding sample standard deviation is 0.3210.
a) Construct a 99% confidence interval for the true average of the difference of residual chlorine
readings between the two methods (assume normality for the difference distribution).
b) Test whether the mean difference is zero at the 5% level.

c) What assumptions did you need to make to carry out this hypothesis test and to determine
the confidence interval? Can you check any or all of these assumptions? If so, how?
Exercise 4
Let 1 and 2 denote true average tread lives for two competing brands of size P205/65R15 radial
tyres. From samples of the two brands we have the data:
n1 = 45, x1 = 42, 500, s1 = 2200, n2 = 45, x2 = 40, 400 and s2 = 1500.
a) Test H0 : 1 = 2 versus Ha : 1 6= 2 at the = 0.05 level.
b) Compute a 95% confidence interval for 1 2 . Does this interval suggest that 1 2 has been
precisely estimated?
c) What assumptions did you need to make to carry out this hypothesis test and to determine the
confidence interval? Can you check any or all of these assumptions? If so, how?
Exercise 5*
Are male university students more easily bored than their female counterparts? The following are
data from a study in which a scale, called the Boredom Proneness Scale, was administered to 97 male
and 148 female students. Does the following data support the research hypothesis that the mean
Boredom Proneness Rating is higher for men than it is for women? Test the appropriate hypothesis
at the 5% level of significance.
Gender sample size sample mean sample sd
Male
97
10.40
4.83
Female
148
9.26
4.68
Exercise 6
Suppose 1 and 2 are the mean stopping distances (in m) at 80 km/h for cars of a certain type
equipped with two different types of braking systems. An expert claims that, on average, cars equipped
with braking system 2 need more than 10 m more to stop than cars equipped with braking system 1.
Assuming normality of the stopping distances, use a two-sample t-test at a 1% level of significance to
test H0 : 1 2 = 10 against Ha : 1 2 < 10. The sample data yield:
n1 = 6, x1 = 115.7, s1 = 5.03, n2 = 6, x2 = 129.3 and s2 = 5.38.
Exercise 7
In the article Estimating the current mean of a process subject to abrupt changes (Technometrics,
37, 311-323), Yashchin (1995) discusses a process for the chemical etching of silicon wafers used in
integrating circuits. This company wishes to detect an increase in the thickness of the silicon oxide
layers because thicker layers require longer etching times. Process specifications state a target value
of 1 micron for the true mean thickness. Historically, the layer thickness has a standard deviation of
0.06 microns. You may assume that layer thickness is normally distributed.
a) A recent random sample of four wafers yielded a sample mean of x = 1.134 microns. Conduct a
hypothesis test to determine whether the true mean thickness has increased. Use a significance
level of 0.05.
b) Using the sample information in a), construct a 95% confidence interval for the true current mean
. Use this interval to determine whether the mean thickness has changed. Discuss the relationship
of the 95% confidence interval and the corresponding hypothesis test.
c) Recall that the power of a test is an expression of its ability to detect when an alternative hypothesis
is true (Slide 78, Lecture 7). The power is a function of the particular value considered under the
alternative hypothesis, and is given by
power = 1 = P(reject H0 when it is false).
Find the power of the test you used in part a) to detect a change in the true mean thickness to
1.01 (i.e if in fact = 1.01). Interpret this value.

d) Find the sample size required to achieve a power of 0.85 when = 1.01.
e) In this question you assumed that layer thickness was normally distributed. How could you check
whether this is a plausible assumption? What else did you assume to do these analyses? Can you
check these assumptions?

End of the tutorial class for Week 11.

CVEN2002/CVEN2702
Engineering Computations - Statistics
Statistics Tutorial Class Week 12 : Regression Analysis
Instructions
The exercises that will be covered during the tutorial class will be chosen from those printed here.
Note that not every topic in the course can be covered by the tutorial exercises. You are expected
to take an active part in the learning process, by working by yourself on most of the topics.
In particular, you are expected to attempt these tutorial questions beforehand. During the tutorial,
the tutor will go through the answers to some of the questions, directing explanation to areas where students
indicate they have difficulty.
Solutions to these exercises will be available from the course web page at a later date.

Exercise 1
In a certain chemical process the reaction time Y (in hours) is known to be related to the temperature
X (inF) in the chamber in which the reaction time takes place according to the simple linear equation
Y = 5.00 0.01X + ,
where is a random disturbance normally distributed with standard deviation = 0.075.
a) What is the true average change in reaction time associated with a 1F increase in temperature?
A 10F increase in temperature?
b) What is the true average reaction time when the temperature is 200F? When the temperature is
250F?
c) What is P(2.4 < Y < 2.6) when X = 250?
d) If an investigator makes five independent experimental runs for a temperature of 250F what is the
probability that all five observed reaction times are between 2.4 and 2.6 hours?
Exercise 2
A regression of Y = calcium content (g/l) on X = dissolved material (mg/cm2 ) was reported in a
1997 article in the Magazine of Concrete Research. The equation of the fitted least squares regression
line was determined :
y(x) = 3.678 + 0.144 x with r2 = 0.860 and n = 23.
a) Interpret the estimated slope 0.144 and the coefficient of determination 0.860.
b) What would be a point estimate of the true average calcium content when the amount of dissolved
material is 50 mg/cm2 ?
c) The value of the total sum of squares was sst = 320.398. Calculate an estimate of the error standard
deviation in the simple linear regression model.
Exercise 3*
Mist (airborne droplets or aerosols) is generated when metal-removing fluids are used in machining
operations to cool and lubricate the tool and work-piece. Mist generation is a concern of the OSHA
(Occupational Safety and Health Administration), which has recently lowered the allowable mist generation standard for the workplace by a substantial amount. In a 2002 article in the Lubrication
Engineering journal the following data was provided on :
X = fluid flow velocity for a 5% soluble oil (cm/sec)
Y = the extent of mist droplets having diameters smaller than 10 mm (mg/m3 )

xi 89
177 189 354 362 442 965
yi 0.40 0.60 0.48 0.66 0.61 0.69 0.99
Various regression output and graphics are included for this question. Use these in your answers to
the following questions :
a) The investigators performed a simple linear regression analysis to relate the two variables. Does a
scatter plot of the data support this strategy?
b) What proportion of observed variation in mist can be attributed to the linear relationship between
velocity and mist?
c) The investigators were particularly interested in the impact on mist of increasing velocity from 100
to 1000 (a factor of 10 corresponding to the difference between the smallest and largest x-values in
the sample). When X increases in this way is there substantial evidence (with = 0.05) that the
true average increase in Y is less than 0.6?
d) Estimate the true average change in mist associated with a 1 cm/sec increase in velocity and do so
in a way that conveys information about precision and reliability.
e) Comment on the normal quantile plot for residuals.
f) Provide a 95% confidence interval for the true average mist when fluid flow velocity is set at 500
cm/sec. Compare this with the graphical display.
g) Provide a 95% prediction interval for a future value of average mist that will be observed when the
fluid flow is set at 500 cm/sec. Compare this with the graphic display.
h) Explain why the interval in part g) is wider than that in part f).
i) OHSAs draft standard requires that the mist exceed 1 mg/m3 at most 2.5% of the time. Use the
attached graph to determine the appropriate level of fluid velocity to meet this requirement.
1

0.9

Extent mist

0.8

0.7

0.6

0.5

0.4
0

100

200

300

400
500
600
Fluid velocity

700

800

900

1000

Linear regression model:

y ~ 1 + x1
Estimated Coefficients:
Estimate
__________
(Intercept)
x1

0.40412
0.00062108

SE
__________

tStat
______

pValue
__________

0.034589
7.5792e-05

11.684
8.1945

8.071e-05
0.00044032

Number of observations: 7, Error degrees of freedom: 5

Root Mean Squared Error: 0.054
R-squared: 0.931, Adjusted R-Squared 0.917
F-statistic vs. constant model: 67.2, p-value = 0.00044

#predicted mean and confidence intervals

>> [ypred,yci] = predict(MyFit,500)
ypred =
0.7147
yci =
0.6562
0.7731
#prediction interval
>> [ypred,yci] = predict(MyFit,500,Prediction,observation) ;yci
yci =
0.5639
0.8654
Thats it for tutorial exercises. You should try to do the rest in MATLAB on your own time.

Exercise 4
The file rain.txt contains rainfall (first column, in mm) and runoff (second column, in mm/h) measurements at Pontelagoscuro on the Po river in northeast Italy, for the 31 years 1918 and 1948. The
data are to be used to produce a model for predicting runoff in terms of rainfall.
a) Recover the matrix rain, imported from the rain.txt data set into Matlab in the first lab class.
Define two vectors rainfall and runoff as the two columns of the matrix rain.
b) Produce a scatter-plot of runoff versus rainfall, and comment on the shape of the relationship
between the two variables. Label appropriately. Write an equation for the regression model that
you would like to fit to these data.
c) Using the Matlab command lsline, add the least squares regression line to the plot. Enter help
lsline if you need more information on this command.
d) Use the Matlab command fitlm to fit a linear regression model
Y = 0 + 1 X +
to the data. Here, the response Y is the runoff and the predictor X is the rainfall.
Note: the fitlm function produces an object which contains just about everything you would want
to know about the model. Some of the information is printed by default, while for others we have
to use functions with the model object to extract information. For this question start by fitting
the model and creating the linear model object using the following code:
>> rainMod=fitlm(rainfall,runoff)
From the produced output, answer the following questions:
(i) What is the equation of the fitted line?
(ii) What is the estimated value of , the standard deviation of the error term ?
(iii) Can you conclude that the rainfall amount has a significant influence on the runoff? Test the
relevant hypothesis at level = 0.05.
(iv) What is the estimated expected change in the runoff for a change in rainfall of 1 mm?
(v) Use the coefCI function to determine a 95% two-sided confidence interval for 1 .
(vi) What proportion of variation in the observed runoff is explained by the variation in the rainfall?

(vii) What is the (sample) correlation between rainfall and runoff?

(viii) Use the function plotResiduals to obtain a plot of the residuals versus the fitted values, and
a normal quantile plot of the residuals, and comment briefly on these plots. Does that validate
the model?
Exercise 5
Milk is an important source of protein. How does the amount of protein in milk from a cow vary with
milk production? The article Metabolites of Nucleic Acids in Bovine Milk (Journal of Diary Science,
1984, 723-728) reported the data contained in the file milk.txt for Holstein Friesian cows.
a) Recover the matrix milk, imported from the milk.txt data set into Matlab in the first lab class.
The first column gives the milk production (in kg/day) for a cow and the second column the
corresponding milk protein production (in kg/day). Define two vectors x and y as the two columns
of the matrix milk.
b) Produce a scatter-plot of protein content (y) versus milk production (x), and comment on the
shape of the relationship between the two variables. Label appropriately. Write an equation for
the regression model that you would like to fit to these data.
c) Using the Matlab command lsline, add the least squares regression line to the plot. Enter help
lsline if you need more information on this command.
d) Use the Matlab command fitlm to fit a linear regression model
Y = 0 + 1 X +
to the data. Here, the response Y is the protein content and the predictor X is the milk production.
e) Write down the equation of the fitted regression line.
f) Does the simple regression model specify a useful relationship between production and protein?
Test at level = 0.05 the significance of the predictor X in the model and draw a conclusion.
g) Estimate the true average protein content for all cows whose production is 30 kg/day. Use a
confidence interval of 99%.
h) Calculate a 99% prediction interval for the protein from a single cow whose production is 30
kg/day.

End of the tutorial class for Week 12. (Graphs on next page)

Linear model output

1.2
1.1
1

Extent mist

0.9
0.8
0.7
0.6
0.5

Data
Fit
Confidence intervals
Prediction intervals

0.4
0.3
0

100

200

300

400

500

600

700

800

900

1000

Fluid velocity

Normal probability plot of residuals

0.9

Probability

0.75

0.5

0.25

0.1
-0.06

-0.04

-0.02

0.02

Residuals

0.04

0.06

0.08

0.1

CVEN2002/CVEN2702
Engineering Computations - Statistics
Statistics Tutorial Class Week 13 : ANOVA
Instructions
The exercises that will be covered during the tutorial class will be chosen from those printed here.
Note that not every topic in the course can be covered by the tutorial exercises. You are expected
to take an active part in the learning process, by working by yourself on most of the topics.
In particular, you are expected to attempt these tutorial questions beforehand. At the tutorial,
the tutor will go through the answers to some of the questions, directing explanation to areas where students
indicate they have difficulty.
Solutions to these exercises will be available from the course web page later.

Exercise 1
An experiment was carried out to compare the flow rates of 6 different types of nozzles, types A, B,
C, D, E and F. ANOVA calculations yielded an observed F value f0 = 4.2 from 66 observations. State
H0 and Ha for the analysis of variance, and carry out the hypothesis test at level = 0.05, giving the
range of values for the p-value which can be determined from the tables, and stating your conclusion
in plain language. What assumptions need to be made for an ANOVA to be an appropriate test? How
could you check if they are plausible assumptions?
Exercise 2
An experiment was carried out to compare the yield of 4 different crops. Random samples were
taken of size 12 from each crop. The observed msEr was 8.2 and the observed value of the test statistic
was f0 = 2.8.
a) Draw up and complete the ANOVA table.
b) Carry out the test using a significance level of = 0.01. Include the statements of H0 and
Ha , rejection criterion, observed value of the test statistic, p-value, and your conclusion in plain
language.
Exercise 3
Having a pet may reduce the owners stress level. To examine this effect, researchers recruited 45
women who said they were dog lovers. The subjects were randomly assigned to three groups of fifteen
women. One group did a stressful task alone; one group with a good friend present; and one group
with their dog present. The heart rate during the task is one measure of the effect of stress. Heart
rates (beats per minute) during stress were recorded for these 45 women. Below you see the computer
ANOVA output for these data.
Analysis of Variance for Beats
Source
DF
SS
MS
Group
2
2387.7
1193.8
Error
42
3561.3
84.8
Total
44
5949.0

Level
Control
Friend
Pet

N
15
15
15

Mean
82.524
91.325
73.483

StDev
9.242
8.341
9.970

F
14.08

P
0.000

Individual 95% CIs For Mean

Based on Pooled StDev
-----+---------+---------+---------+(-----*-----)
(-----*-----)
(-----*-----)
-----+---------+---------+---------+-

Pooled StDev =

9.208

72.0

80.0

88.0

96.0

a) What are the null and alternative hypothesis being tested?

b) From the ANOVA output find the observed value of the F -statistic and the corresponding p-value.
State both degrees of freedom for the F -statistic.
c) Briefly describe your conclusions.
d) What are the sample means for the heartrates of the three groups of women in the study?
e) There are three groups considered in this study. If you carry out t-tests to compare them two at a
time (pairwise t-tests), how many t-tests are there? Using the Bonferroni adjustment, carry out a t
test comparing the group which did a stressful task alone and the group with a good friend present.
Exercise 4
A university development director conducted a survey of starting salaries offered to education majors in three states. Ten offers were recorded for each state, and a partially completed ANOVA table
appears below.
Source
Treatment
Error
Total

df
SS
MS
F
A 17470016
B
E
27
D
874482.48
C 41081016

a) Complete the ANOVA table by determining the values for A, B, C, D, E.

b) Test to see if there is a significant difference in mean starting salaries between states at level
= 0.05. Clearly state the null and alternative hypothesis both in mathematical terms and in
simple language.

End of the tutorial class for Week 13.