Resources to support the learning and teaching of Mathematics, Statistics and Operational Research in Higher Education

Maths, Stats & OR Network, The University of Birmingham, Edgbaston, Birmingham, B15 2TT.
Phone: +44 121 414 7095 Email: info@mathstore.ac.uk Web: www.mathstore.ac.uk
Flexible Learning
in Statistics
Clive Anderson, University of Sheffield
Andrew Bertie, The Open University
Steve Bond, London School of Hygiene & Tropical Medicine
Jane Bruce, London School of Hygiene & Tropical Medicine
Caitlin Buck, University of Sheffield
Nick Fieller, University of Sheffield
Barbara R Smith, University of Southampton
Jane Williams, The Open University
Flexible Learning in Statistics
Clive Anderson
University of Sheffield
c.w.anderson@sheffield.ac.uk
Andrew Bertie
The Open University
a.bertie@open.ac.uk
Steve Bond
London School of Hygiene & Tropical Medicine
steve.bond@lshtm.ac.uk
Jane Bruce
London School of Hygiene & Tropical Medicine
jane.bruce@lshtm.ac.uk
Caitlin Buck
University of Sheffield
c.e.buck@sheffield.ac.uk
Nick Fieller
University of Sheffield
N.Fieller@sheffield.ac.uk
Barbara R Smith
University of Southampton
bs99@socsci.soton.ac.uk
Jane Williams
The Open University
j.e.williams@open.ac.uk
2
Flexible Learning in Statistics
Contents
Adrian Bowman
Introduction
Barbara R Smith
Tutorial Tennis
Steve Bond & Jane Bruce
Using Learning Technology to Deliver a
Distance-based MSc in Epidemiology
Clive Anderson, Caitlin Buck & Nick Fieller
The Sheffield Statistics MSc Goes DL
Jane Williams
Distance Teaching of Statistics to
Undergraduates at the Open University
Andrew Bertie
Java Applications for Teaching Statistics
3
4
9
16
29
33
3
Introduction: Adrian Bowman, University of Glasgow
This workshop was held in conjunction with the Statistical Computing Section of
the Royal Statistical Society and the Open University. The title of the workshop
was Flexible Learning. This raises the question of what is meant by the word
flexible. Is it flexible in geography, which is expressed in the alternative phrase
distance learning? Is it flexible in time, so that students from the same geographical
location have the opportunity to work around other commitments? Is it flexible in
learning style, so that teaching is conducted in a manner which enables those
who learn in different ways to handle the course? Is it flexible in background,
recognising the different levels of preparation which maybe present in the class.
It was useful to distinguish these, as discussions often mix them together in the
context of IT, distance and open learning.
The traditional approach to teaching in higher education is through lectures, tutorials
and labs, but these can now be augmented by a whole host of other support
mechanisms. This might include lecture notes on the web, an electronic glossary,
various forms of on-line interaction amongst students and or teachers, the use of
dynamic graphics to explain concepts, quizzes for assessment and problem solving
material for higher level forms of learning. If flexible learning means that the
traditional structure of lectures, tutorials and labs is removed then there are
substantial issues of course organisation and it was extremely interesting to see
how these were tackled by the speakers in the workshop.
The word integration seems a key one. Integration of design is important so that
the course hangs together as a coherent set of resources and activities addressing
important aims. Equally important is integration of delivery so that the different
resources can be accessed in an easy and flexible manner. The idea of a virtual
learning environment is now well established and some of the talks at the workshop
gave interesting insights on how these might be used. From a more statistical
perspective integration of analysis and other types of software is also important
and the most promising vehicle for achieving this currently seems to be R.
Some clear themes emerged from the contributions at the workshop. One of
these is the very substantial cost of developing material in this area. Another is
the need to maintain the material and a third is the possibility of sharing developed
material so that others need not reinvent what already exists. Possible ways of
addressing these issues are through collaborative projects, the development of
common standards and a swapshop or library system where material can be shared.
We would be delighted if the website http://ltsn.mathstore.gla.ac.uk could be
used for this purpose and we would encourage everyone to consult it to view the
material already there and to make further contributions.
4
In the early 1980s a taught course was offered at a local Southampton college
with the aim of preparing students for entry to the examinations of the Institute
of Statisticians. There was a dearth of students willing to commit themselves
to part-time study and so the courses were withdrawn. However, we sensed
that there was a demand particularly from mature students to improve their
statistical knowledge. So we began to convert the classroom notes into a
number of self-contained study guides with exercises based on the learning
material.
Now that the students could learn at a distance, their numbers began to grow
locally, nationally and internationally. The then Institute of Statisticians was
aware of our courses, and introduced prospective students to us. Marketing
was easy! Our venture into distance learning started in the early 1990s when
the courses were transferred to the Department of Social Statistics. So where
are we in 2002? What do we do and who are our students? And whats it got to
do with tennis?
Stages in Student Enrolment
(a) The Welcome Booklet: The game of tennis begins when a prospective student
shows interest in our courses and makes contact by telephone, fax, e-mail or
letter. We have a link on the Departmental website and also printed brochures
giving details of the courses.
Those who make this initial contact have various backgrounds ranging from an
absolute beginner who could be a clerk in a statistical department or someone
who needs to change direction at work having previously graduated with a
degree in Economics or even in Classics. Once a suitable level has been
determined, the student completes an enrolment form and pays a course fee
and in return is sent our Welcome booklet. This describes how the course
proceeds and what is expected of the student.
Tutorial Tennis
Distance Learning in Preparation for the
Examinations of the Royal Statistical Society
Barbara R Smith, University of Southampton
5
And so the game of tennis has begun! The first serve was the request for
information. The ball was returned as a request to complete the enrolment
form and for the student to introduce oneself to the tutor. The Welcome
booklet was sent in response to the enrolment. The interplay between student
and tutor resembles further volleys.
On the last page of the Welcome booklet is a plan of the course and students
are invited to add the dates when they plan to submit their assignments for
marking.
Knowing that our students are busy and mature people, it is essential that they
are not over-ambitious and can fit their study periods around other important
events in their lives. Many will have deadlines to meet at work and some will
know that, for example, March is concerned more with the end of the financial
year than with study and that they could not concentrate whole-heartedly on
their studies at that time. Others will have planned family holidays, or even
have planned for a family! Students are asked to look at their diaries and to
add sensible dates to the plan of the course. No two students will have the
same set of dates and from this point on each is considered as an individual
rather than as a class member. This also means that we enrol students at any
time of the year.
(b) Receiving the first Study Guide: As the game of tennis proceeds, the receipt
of the provisional time-table by the tutor is a signal for the first two of the thirty-
five study guides to be sent by post. Students are asked to start work on the first
book. Sending two books ensures that they always have some work in hand in
case the tutor is slow to respond or there is a postal delay. They are reminded
that the material they receive is just a guide and needs to be digested in small
amounts and studied in sequence.
Content of a Study Guide
Each study guide contains three parts which are colour-coded for easy
identification. The first part offers explanations of statistical topics. Another
part provides examples of examination-type questions. Suggested solutions for
these Answer problems are placed in a sealed envelope and sent inside the
study guide. Students are asked to try to solve the problem first and then to
open the envelope! The third part consists of Tutor examples that are often
taken from past examination papers. No solutions are offered because the tutor
needs to see how the student has tackled each problem and give guidance
when some reassurance is needed. The tutor also offers comments on
presentation and interpretation.
6
In a game of tennis, the ball can go wide or hit the net and this interrupts the
flow of the return. Similarly, the pattern can be broken for our students. When
it is clear to the tutor that a student has not understood a particular point, new
notes are created and sent to the students and if the modified explanation is
found helpful, then those notes are incorporated into future editions of the study
guide. Hence the presentation and content of the learning/teaching material
are being continually updated.
The Subject Matter
Today, we prepare students for the examinations of the Royal Statistical Society.
The syllabuses for these examinations form the basis of the subject matter in
our study guides. The examinations are at three levels: Ordinary Certificate,
Higher Certificate and Graduate Diploma (with four core papers and one Option
paper).
Details of the objectives and syllabuses may be obtained from the RSS.
The official objectives of the RSS examinations, are in turn, the objectives of
our courses. On a broader front, we hope to give our students the opportunity
to continue their professional training.
The Students Responsibilities
You cannot play tennis on your own! Students have to be active and relate to
their tutor. All students will read the same printed words but their responses
will be personalised as they read their favourite text book, make their own
notes and continue the communication exercise by contacting their tutor
whenever they need help. They should not be afraid of making a nuisance of
themselves their tutor is available for them individually.
We are aware of the sense of isolation that Distance Learning students often
experience. One way to alleviate this is through the network where two or
more students working at a similar level can communicate with each other.
Another is through our Winter School which runs for three days each January.
The aim is to bring students into the University environment from their places
of work and to introduce them to each other and to their tutors. Also, if students
live near Southampton and need individual help, then they are encouraged to
meet with their tutor at the University. From time to time, an employer like the
ONS or DEFRA will invite the tutor to visit the students at work for a one or two
day tutorial.
7
Our Current Students
At the moment our students are spread worldwide. Nearly all are employed
and many have their course fees paid by their employer even if no study leave
is granted. Some students do pay their own fees. We have offered bursaries to
a small number of these students. Apart from students who are resting and
who may suddenly wake up, we have 56 students on our books today.
How do we view our strengths and weaknesses? - Perceived Strengths and
Weaknesses
By giving students a quick turn round on their marked assignments students feel
that their attempts are important and this encourages the keen ones to make
further effort. Here the game of tennis moves with real enthusiasm. There are
some students who for various reasons such as illness, busyness or family pressures
or loss of interest appear to fall by the wayside. Sometimes a student appears
to have dropped out and then I receive a message that says: Do you remember
me? Can I pick up where I left off one or two years ago? Our version of
Distance Learning encourages such flexibility. We do not hurry them although
lax students will occasionally receive a reminder asking them if they are
experiencing problems with a particular topic and to contact their tutor so that
they can be re-directed.
Sixteen per cent of this years students moved from one study guide to the next
in less than one month, but some of the African students who were out on field
work took 6/9 months between submitting assignments. This very slow game
of tennis has not helped their progress! Certain students lack motivation often
because their employer offers no reward for success in examinations.
From this you will see that the one-to-one contact between a student and a
tutor forms a relationship similar to that of two tennis players, where each is
alert to the movement of the other and responding as necessary.
Summary and Recommendation
Our courses have the following characteristics:
We have no need to advertise students are looking for a course.
The syllabuses are broad in nature but the Option paper at the Diploma
level gives depth.
Students are responsible for their own learning and work at their own pace.
8
Some students are unable to gauge the amount of time required for study.
Interest is maintained by having one study guide in hand.
It is essential that work is submitted for marking and feedback.
Study guides are being constantly edited in the light of student responses.
The RSS has three levels. Candidates may terminate their studies after any one
level and be awarded, after suitable practical experience, with the appropriate
qualification. We do encourage them to continue to the next level but many
are content with an intermediate qualification because it is sufficient for their
needs, time or abilities. Many students as statistical consumers need to know
enough Statistics to be able to communicate with statisticians who have attained
the Graduate Diploma.
In our experience the most difficult transition for students is from the non-
calculus based Ordinary Certificate to the calculus-based Higher Certificate.
This requires a considerable investment in learning new mathematical skills
which a large proportion of students find daunting, yet they are still keen to
extend their statistical knowledge. Perhaps a new terminal route to a non-
calculus based Advanced Ordinary Certificate might be considered. This
could expose students to statistical ideas beyond that of the Ordinary Certificate
but without the need to make the investment in improving their mathematical
skills first. This could be a valuable terminal qualification and perhaps also
increase the longer term recruitment to the Higher Certificate.
Ordinary
Certificate
Higher
Certificate
Advanced
Ordinary
Certificate
Graduate
Diploma



9
Using learning technology to deliver a
distance-based MSc in Epidemiology
Steve Bond and Jane Bruce
London School of Hygiene & Tropical Medicine
Epidemiology: Principles and Practice (EPP) is a Postgraduate Diploma and
Masters course offered by the London School of Hygiene and Tropical Medicine
(LSHTM). The course materials consist largely of computer-assisted learning
(CAL) on CD-ROM, together with some paper-based units, workbooks and
readers.
The course has been running almost 3 years, with a total intake of about 300
students. To date 9 students have already graduated from the MSc, with a
further 20 expected this year. Our students come from 38 different countries,
with about 30% based in the UK.
Tutorial support is provided by email and web-based conferencing.
The course was developed and is run by LSHTM, but delivered via the University
of Londons External Programme. It is the first distance-based Masters course
within the External Programme to use CAL materials.
Course structure
The course has 4 compulsory core study units, totalling 600 study hours. A
student completing these can sit the Postgraduate Diploma, or may continue to
the full MSc.
The core units are as follows:
Fundamentals of epidemiology
Statistics with computing
Practical epidemiology
Research planning & scientific reporting
Of these, the first two are computer-based and the other two paper-based. To
complete the MSc, the students must complete 6 advanced study units, making
up a further 600 hours of study. There are 2 compulsory units, using a combination
of computer and paper based materials:
10
Study design: Writing a grant application
Statistical methods in epidemiology
The other 4 units are chosen from the following:
Epidemiology of communicable diseases
Epidemiology of non-communicable diseases
Advanced statistical methods in epidemiology
Genetic & molecular epidemiology
Additional units from the other 2 paper-based DL courses
(Infectious Diseases & Health Systems Management)
The students have 2-5 years to complete the MSc or 1-5 years for the Postgraduate
Diploma.
Study materials
Each study unit is comprised of sessions, each of which is equivalent to a
lecture in face-to-face teaching. These sessions are either computer-based or
paper-based. Each study unit may be accompanied a workbook of practical
exercises, and a reader and textbooks to which the session will refer. There are
also a number of tutor-marked assignments (TMAs) that the student can complete
and return to LSHTM for marking (these are not formally assessed).
All sessions have learning objectives, clearly stating what the learner should
expect to be able to do after completing the session. The text was written
using clear and simple English (since many of our students are not native English-
speakers), in a familiar, non-intimidatory style, also taking into account cultural
differences.
Each study unit contains all the learning materials required to achieve the
learning objectives. As such, the study units form comprehensive self-tutorials,
but as we shall see later, students also have recourse to email and conferencing
for further support.
Use of CAL
The aim of making certain study units computer-based was to exploit the
interactivity that this approach allows. We included exercises to test students
understanding of the concepts, with discursive feedbacks that explain why
they got the correct or wrong answer. The CAL also allowed for flexible
navigation within and between sessions and study units.
11
Figure 1 shows a screen from one of the Statistics with Computing sessions.
The student can manipulate the diagram on the right by varying the mean and
standard deviation. In this way they can explore the behaviour of the Normal
distribution, and use their findings to complete the exercise on the left. In the
exercise they choose the missing word from a list of options, and receive an
explanation of why that is the right or wrong choice.
It is not only a long distance affair
Although developed for the purposes of distance learning, the materials have
also been used in a face-to-face setting. The computer-based sessions are all
available on the LSHTM internal network, where they are used as supplementary
or replacement materials on certain courses. They are also used extensively
by students for revision. The modular nature of the course means that individual
sessions can be used to target specific areas that students find difficult.
Use of conferencing
Another important aspect of the course apart from the course materials
themselves is the use of online conferencing. This allows students to express
their understanding of a concept and receive feedback on which to base a re-
evaluation of that understanding. This allows for a much richer form of interaction
than that which is possible between a student and a computer program.
Figure 1
12
Conferencing provides a way for students to communicate with their tutors and
with their peers, providing a number of benefits. Importantly, it counters the
isolation of distance learning by putting students in touch with others who
share the same difficulties and frustrations. It provides a medium for tutorials,
seminars, and other academic discussion. It allows all students to benefit from
a question asked by a single student. Finally, it creates an online community
of practice, of people working together to further their common understanding
of a subject.
Choice of conferencing system
We had certain constraints when choosing a conferencing system. Through an
initial survey of LSHTM alumni in 1998, we ascertained that our target audience
did not have affordable & reliable web access, though almost all had access to
email. Therefore we needed a system that would provide the benefits of online
conferencing while still allowing participation by email.
At the time the only solution on offer was WebBoard by OReilly (now Avika).
This is still in use on the course today, but it is now 4 years old and continuing
reliability problems and difficulties in administering the system mean that it is
time to consider upgrading or changing to a different system.
The WebBoard interface
Figure 2
Figure 2 shows the WebBoard interface, as seen by a student who has web
access. The conferences are listed on the left, showing a range of different
13
topics covered. On the right are the messages that have been posted to the
opened conference. The system threads messages together by topic, so that
discussions can be read naturally as they develop.
Email-only users are able to post messages to these conferences by sending
email to a special address, and they receive a daily digest of all the messages
posted in each conference. In this way they do not miss out on any of the
discussion, though the fact that they receive the emails chronologically, rather
than in a threaded form, means that it is less easy to follow the common themes.
Types of conference
Study unit-specific: In Figure 2 you can see that many of the conferences relate
to the study units (EP101, EP102) that make up the course. In these conferences
students discuss the issues that arise as they work through the study materials.
Common rooms: In the common rooms students are free to discuss anything
they want, whether it be about the course or just general chit-chat. This helps
to build up a sense of community amongst the students. Certain common
rooms are restricted; for example, the Student Caf is accessible only by
students, so they can ask questions and discuss the course without worrying
about what a tutor might think. Similarly, the staff also have their Staff Room,
which is only accessible to tutors.
Seminar rooms: In the Seminar Room, we invite leading researchers from
within LSHTM to come online and answer questions from students on the subject
of their research speciality. We also have a Virtual Lecture Theatre; here,
lecture notes are provided for a forthcoming lecture at LSHTM, and the EPP
students can post questions that they would like to ask the lecturer. These
questions are read out in the lecture itself, and the lecturers response recorded
and passed back to the students. The aim is to allow DL students to participate
in some of the day-to-day life of the School.
Example: Outbreak investigation
Our outbreak investigation exercise is an example of the CAL software combining
directly with the conferencing system. Students are assigned to small groups,
to investigate a simulated outbreak of cholera in a Gambian village. In this
task they use a program called Mappa, which allows them to map the progress
of the outbreak on a plan of the village. The students can export their results
from this program, and collaborate and exchange data via the conferences.
Figure 3 shows an example of some work done with Mappa.
14
Level of use of conferences
After the first years use of the conferencing system, we analysed the logs to
see how widely it was being used. We observed a 4-way split in modes of use,
into the following groups:
Active users: those who posted more than 2 messages
Transient users: those who posted only 1 or 2 messages
Passive users: those who chose to be registered but never posted
Absent users: those who never responded to our invitation to register
The proportions in each group after the first year were 50%: 15% : 27% : 8%.
This was fairly encouraging, because in online courses (where participation is
not compulsory) an active of proportion of 33% is more usual. Nevertheless we
wanted to encourage the other 50% to become active in the second year.
To encourage absent users to register we sent reminder emails explaining the
benefits of the conferences. For passive users, we sent simplified instructions
(in case they were having difficulty in posting), and made suggestions on what
to post. We also introduced new attractions, such as the Virtual Lecture Theatre,
to encourage participation.
However, the latest figures (now incorporating 3 years intake of students, from
Oct 1999 to May 2002) show a split of 43% : 13% : 35% : 9%, despite the
measures described above. There has been evidence from other sources to
indicate that our first years intake were more familiar and enthusiastic in general
Figure 3
15
with computer-assisted learning, so it may be that they were always more likely
to be active than those that joined later.
Costs
Our initial budget for development was a loan of 300,000 from the University
of London, but that was certainly exceeded by the time the final study unit was
completed. We estimate that the development required something in the range
of 10-14 person-years, over an actual period of 4 years.
Operational costs have also been higher than expected. The tutoring
requirement during the second year was about 20-30 person-hours per week,
but now in the third year we have an extra 80-90 students, and so the tutoring
requirements increase accordingly. Some of this tutoring load is allocated to
staff already teaching in LSHTM, but tutors were also employed specifically to
support DL. There is also the cost of technical support, which at the moment
occupies a 50% full-time equivalent post.
Overall, operational costs break down into 50% teaching & support, and 50%
to cover materials, copyright, admissions and exams.
Evaluation & feedback
We conduct student evaluations at the end of each year, but so far the response
rate has been too low to generate useful results. However, the qualitative
feedback throughout the year has been largely positive:
Students described the computer-based sessions as exciting and
stimulating
Tutor feedback on conferences was considered very helpful
Students stated that they felt part of a community of students
On the negative side, there were some common themes:
Students complained about the number of mistakes in the early versions of
the computer-based sessions, which arose due to poor quality control
Some students found the volume of emails coming from the conferences
overwhelming
The EPP course website is at www.lshtm.ac.uk/cal/epp, where you can find out
more about the course, and download the prospectus or the admission form.
The authors are no longer working on the EPP project, but specific enquiries
(other than admissions) can be directed to the distance learning co-ordinator
for LSHTM John Ackers, email John.Ackers@lshtm.ac.uk
16
The Sheffield Statistics MSc Goes DL
Clive Anderson, Caitlin Buck and Nick Fieller
University of Sheffield
The University of Sheffield MSc in Statistics started in 1965. Until now it has
been given entirely in a traditional format (with lectures, laboratories and
exercise classes) and exclusively to students able to attend the University.
Most were full-time students but occasionally we had part-time students taking
the course over two or three years. Since we generally made no particular
provision for these in constructing the timetable most part-timers were not
working in regular jobs. From this year, however, we are offering a new part-
time distance learning version of the course alongside the established version.
This is available to students studying anywhere in the country and we anticipate
that most students taking it will be in full-time employment. This note explains
the motivation for the new course, outlines how it is being implemented through
web-based and other means and offers some reflections on our experience so
far.
Background and Motivation
The Sheffield MSc is broad-based, aiming to give a thorough preparation for a
professional statistical career in any branch of the subject. Over the years
since 1965 well over 400 students have graduated from the course. Those
looking for jobs have had little difficulty in finding them and many others have
continued with postgraduate study for a PhD, sometimes after a year or two of
working. Typically a third and sometimes more than a half of each MSc cohort
eventually complete a PhD either in Sheffield or elsewhere; those going straight
to employment often work in research teams either as RAs in academic
environments or within R&D groups in large companies. The result is that
Sheffield MSc graduates are well represented throughout the statistical profession
both in the UK and elsewhere. At any statistical meeting in the UK of more
than a handful of people there is usually likely to be at least one Sheffield MSc
graduate, indeed there were three at the meeting on the 9th May when this
presentation was given at the OU.
Demand for the course continues to be healthy. In the year 20002001,
immediately before the developments described here, there were 17 students,
in line with the average complement over recent years. The course can thus
be counted as successful in helping supply a national need for statistical training.
Why then does it need a distance learning capability?
17
A bid in 2000 to EPSRC for award of a Masters Training Package prompted the
Department to try to envisage the likely pattern of needs for statistical training
over the next decade or so. The background to the thinking was concern about
the continuing national shortage of statistically trained people, as evidenced
by widely-reported difficulties in filling statistical jobs and the national lack of
understanding of statistical ideas. It is interesting that exactly the same concern
lay behind the setting up of the MSc course in Sheffield in 1965. The recently
changed arrangements for support of undergraduate students have resulted in
many students completing their first degree course with a large burden of debt.
It is understandable that many wish to take a job rather than a Masters course.
A change in the pattern of recruitment to traditional MSc courses might be
expected in consequence. When, later, the student feels able to contemplate
MSc training they may be established in a job and may have personal and
family commitments that would make full-time attendance for a traditional
MSc course difficult. For such a student, it was argued, some more flexible,
part-time arrangements would be helpful. Such arrangements could be attractive
to other groups too: those trained in other disciplines but having to do statistical
jobs with only limited background, and those trained in statistics some years
ago and now wishing to update skills and knowledge. Accordingly, the
Department of Probability & Statistics decided to set up a part-time distance
learning arm to the Sheffield MSc.
There was unanimous agreement that the new course should be of high quality
fully equivalent in standard to the current residential MSc. We see the only
way to achieve equivalence is to have identical assessment procedures
(examinations, assignments and deadlines) for all students taking the course,
whether as a resident or as a distant student. The only difference would be that
the distant students would take the course over two or maybe three years instead
of twelve months. A practical necessity was that the new course would be
developed and sustained within available departmental resources.
The MTP proposal including the suggested distance learning element was
successful and brought half-time funding for a Distance Learning Manager
(DLM), and some secretarial help. These funds have been used to buy out
teaching and administration time of an additional regular member of the
lecturing staff. The DLMs job is to coordinate development and delivery of
the materials for the new course and to liaise with the distant students. Apart
from this extra support, the set-up and maintenance costs for the new course
have had to come from within the department. We comment on resource
issues below. The other elements of the MTP funding cover support (fees for all
EU students plus a standard maintenance grant for UK residents) for
approximately the equivalent of twelve full time students. However, our intention
stated in our bid is that we will split this to give support to nine residential full-
18
time students and thus cover fees and other expenses (see below) of cohorts of
six DL students per year.
The limits on time and manpower available for development of the new course
dictated evolution rather than revolution. It was decided from the outset that
what would be offered to DL students would be based on existing modules from
the full-time course, rather than completely newly-developed material. It was
also decided that we would use tested methods of delivery rather than attempt
radical innovation. After all, the purpose of the course (and our expertise)
centres on the statistical training it conveys, and not on experimentation in
educational technology. For advice on educational technology and on distance
learning issues we were fortunate indeed in being able to call on the expertise
of a specialist unit within the University, the Sheffield University Learning
Media Unit (LeMU). Their guidance and support have been invaluable.
Following the advice of Adrian Powell and Ruth Sharratt of LeMU we decided
to use the package WebCT for delivery of the course. There may well be
alternatives to WebCT but the advantages of being able to tap into in-house
expertise far outweigh any merits of competitors.
Current Structure
Our MSc course is designed to cover the three elements of technical knowledge
of the core of Statistics, practical & professional skills and work on an extended
project leading to a dissertation. Our aim is to deliver all of these elements in
full without compromise to distant learners.
The existing residential course combines study of these elements by:
Lecture modules, problem solving classes and extended assignments on
various technical topics in Statistics. Typical lecture modules cover
Linear Modelling, Inference, Bayesian Statistics, Dependent Data,
Sampling, Experimental Design, Medical Statistics, Probability Modelling,
Industrial Statistics, Econometrics, Mathematical Biology, some being in
parallel with our level 3/4 MMath undergraduate programme and only the
first few being compulsory for all students;
Modules developing practical & professional skills, learnt through
projects, practice in communication, both written and oral, rle play,
group work and work in the statistical computing laboratory;
A substantial statistical problem, studied individually under the active
supervision of a member of staff and written up as a dissertation
accounting for one third of the overall assessment for the course. Most
work on this would be done over the summer months and often this would
be based on a practical problem brought by a client from outside the
department or even outside the university.
19
Additionally, as a preliminary we distribute full notes and exercises of a
Foundation Block in early summer before the start of the course. This block is
intended to fill in the gaps in basic material; students are expected to submit
work on the exercises before the start of the course and this is followed by some
classes in the preliminary week before the term starts. Other sessions in the
preliminary week are devoted to statistical computing etc.
The only compromise we are making for the DL version initially is that we do
not intend to offer any optional modules in the technical core and so we can
limit the amount of material that has to be prepared and presented in distance
learning form to just the first eight of the topics listed above. The substantial
staff cost per student, especially but not exclusively on the third element, means
that we do not envisage opening the course to large numbers of DL students,
unlike some other distance learning courses in similar areas. It is likely that in
the first few years numbers on the course will be limited to fifteen residential
and about eight (or maybe ten) new DL students per year, two thirds of these
receiving EPSRC support through the MTP funding.
Advice Received
Initial discussions we had with LeMU and within the department and colleagues
elsewhere identified many key needs of DL students which perhaps had not
seemed so apparent for our traditional full-time residential students. However,
we realise now that many of these are relevant to them as well, even if in
lesser degree, and we feel that concentration on these has improved the course
we offer in total.
The first of these concern the qualities of the students themselves. They need
time, determination and commitment to the course to a much greater extent
than the traditional full-time MSc student who proceeds direct to the course
from a first degree. It is not an easy option to take the course part-time as a DL
student and our admissions criteria have to be at least as stringent, if not more,
than for full-time residential students, i.e. a good first degree in a mainstream
mathematical subject with at least some knowledge of basic probability and
statistics. We suspect that in the past there has been the temptation to regard
part-time (though residential) study as an option for borderline candidates.
This year we have been fortunate in attracting well-qualified and highly
motivated students to the course (and resolute in not accepting others) and so
we have not experienced the high drop-out rate that we had been warned to
anticipate at least not yet! Indeed only one of our initial intake of seven has
suspended studies for twelve months because of unexpected domestic disruption
with their employment etc.
20
An element emphasized by many of those offering advice was that for the DL
students we needed to replace the benefits full-time students obtain just from
being resident within the department and attending lectures. In particular these
include the general encouragement of contact with other students and lecturing
staff (and secretarial staff) giving quick resolution of queries and the opportunity
for raising points of discussion as well as the recognition that they are not alone
in finding some technical points in lectures opaque and badly explained. It
was stressed that it was essential to have a definite route map stating in advance
what would be provided, when it would be provided and how it would be
provided, what to expect and what not to expect. Perhaps we have not been
alone in not considering these issues fully for our regular students.
An issue related to that above is the need for rapid and informative feedback,
another point emphasized by our professional educational advisors. Again, we
admit that we had on occasion been lax in providing this for our regular students
who could always be reassured by a quick verbal comment face to face before
the full written comments on submitted work were available. There is not this
opportunity with DL students who need the encouragement that they are on the
right track and not straying off course. It soon became part of the DLMs job to
encourage those of us teaching the modules to provide this quickly.
With the principle of evolution rather than revolution our starting point for the
DL material was existing lecture material, which for most modules consisted of
full printed notes, printed exercises and example sheets and (mostly) handwritten
outline solutions. It was clear that we needed to amplify and supplement the
lecture material to make it more self-contained. More importantly we saw that
we needed to provide more simple tasks (or five-finger exercises) which would
reinforce each lecture and provide further encouragement and self-feedback to
the DL students in particular. These needed to be much simpler and quicker
than the more substantial exercises which in many cases are [parts of] former
examination questions. They had to be simple tests specifically on the material
in just the previous lecture or two and which could be done in a few minutes,
perhaps even just mentally. This was indeed a point that our various advisors
highlighted. Again, this is a feature which is of clear educational benefit to all
of our students, not just the DLs.
Initial Plans
The structure we settled on for the DL students was to have a total of three
residential weeks during their period with us. These would be in September at
the very start of the course and then two further weeks in June in the first and
second years. The first of these covered the foundation block and basic material
but the main purpose of the week was to establish contact between the DL and
21
residential students and between them and the staff, especially those most
closely in the DL provision. The plan included several social events, including
a couple in the evening, some of which were just for the DL students and staff.
The second residential week in June was scheduled to cover final revision
sessions and then the examinations, with the remainder of the week devoted to
various modules including professional skills and group work involving
collaboration between DL and residential students. Again, several social events
were included in the schedule. A similar schedule in June in the second year
was envisaged. DL students spreading the course over three years would
additionally need to come for just the examinations on the final two modules
in the third year.
These residential weeks were specifically designed to establish early contact
and rapport, primarily between the DL students and the staff, the contact with
the residential students was also seen as important but to a lesser degree. To
maintain contact through the year we planned two further strands. The first
was one of the key roles of the DLM. Each DL student was required to send an
informal report to the DLM each week saying how work had progressed over
the preceding days, either by email, and the DLM would chase these up if they
did not arrive. Additionally, the DLM had the rle of personal tutor for all of
the DL students, a rle which was intended to be much more active than the
underwriting safety net provided by tutors of the residential students and so
might involve telephone conversations from time to time. The aim of this
provision was to identify difficulties early and remedy them quickly if possible.
Often these contacts would reveal that problems arose on our side, glitches in
the provision of material and slackness in feedback for example.
The second strand was to take advantage of WebCTs provision of discussion
boards (many similar packages provide such a facility). Some illustrations of
these are given in sections below. As well as separate boards for each module
we established ones for general matters and queries. All staff and students
(including the residents) had access to all of these (students only if they were
taking that module). We included two further boards for separate private
discussion for staff and for students. A key purpose of the separate module
discussion boards was to maintain contact between the lecturer responsible for
the module and the DL students. It was also intended that these should encourage
contact between the DL and residential students taking the same module. Often
these boards would be used for raising technical queries. The lecturer concerned
would monitor these and encourage discussion between the students (both DL
and residents) and provide an answer if no one else did. Further, each lecturer
would provide a summary each week of what material had been covered during
the lectures, what further points had arisen (e.g. if queries had arisen in lectures),
what background material would be needed in the near future and draw attention
22
to any changes or corrections made in material already distributed.
All of the above elements were designed to substitute for the missing experience
of being at lectures and day to day face to face contact between staff and
students. Of course, for the resident students they still had this so these elements
were an additional facility for them though they had to be encouraged to utilise
and, more importantly, contribute to them.
Great effort was put into to enhancing the existing course material. This had
been prepared in either LaTeX or in Word, in about equal proportions, following
the taste of the lecturers concerned. Initial experimentation with converting
everything into HTML quickly revealed that this was impractical. The solution
adopted was to convert both LaTeX and Word into Adobe Acrobat .pdf format
before placing on the WebCT server. Some elements, such as study guides,
syllabuses etc could be converted into HTML and this was preferred where
practical, not least for the sake of speed of access. It may be noted that both
TeX and Word can accommodate hyperlinks which can be carried through into
.pdf format. Had the material been written totally from scratch then HTML
could have been a preferable option.
WebCT provides a calendar facility which can be used as a basis for a Route
Map, but needs to be supplemented by a more discursive study guide to give
an idea of relative proportions of time to spend on different aspects of modules
etc.
Lessons from Experience
Now that the first two teaching sessions have passed it is possible to reflect on
some of the lessons learnt so far. In summary, providing a DL course in parallel
to the existing course has been substantially more work than anticipated.
Opinions differ but most of those of us involved would concur with a figure of
something like an additional 150% load. Roughly, this is composed of about
50% in setup costs and 100% in running the course. Further, much of the
additional load in running it is unscheduled. Overall, most of us feel that the
delivered product (from the point of view of the residential students) has been
enhanced though one drawback noticed by some of us is the danger of over-
teaching of the residential students.
The setup costs consist of enhancing the course material, making it more self-
contained and not reliant, for example, on clarifications and additional examples
that were given in lectures. This cost will eventually reduce but only slowly
none of us feel that we have this right at first iteration. Running the course in
DL mode requires responding quickly to queries and providing rapid feedback.
23
This is essential to prevent DL students in particular from loosing touch (or
feeling that they are loosing touch). Further, it takes much more time to
provide a written reply on WebCT discussion boards than answering someone
verbally. To ensure involvement of all of the students we pressed even the
residential students to use WebCT as the preferred means of raising queries.
On occasions this meant cutting and pasting direct email enquiries onto WebCT
(see below for an example) or on us reporting queries which had been asked
face to face. This was necessary to include the DL students into the community
of people taking the module. However, it requires much more time and effort
than answering quickly face to face with an individual questioner.
A further cause of the substantially increased load was the collection of work
(e.g. exercises on modules) to be marked and returned. Previously this had
been largely optional , both for the lecturer in offering it and for the student in
doing it. This past year we have been much more insistent in requiring the DL
students to submit these exercises since this was a good (and almost only)
indicator for monitoring their progress and contact with the course, unlike
residential students where we can see if they are around and in lectures. Some
of us feel that seeing if students are sitting in lectures is not enough and we
should be harmonising the way all students are monitored.
Provision and monitoring of exercises is perhaps a relatively minor element.
Much of our MSc course is reliant on more substantial assignments and writing
of reports on data analysis. Whereas general comments, and indeed individual
ones, could be conveyed verbally, for DL students there needs to be a more
detailed written report, either on individual submissions or on WebCT discussion
boards.
The provision of all the additional features required by the DL students has
undoubtedly enhanced the overall quality of our MSc course. It could certainly
be argued that this is purely because we have started thinking about educational
aspects (routemap, feedback, simple reinforcing task sheets etc) which we
could and should have done irrespective of the DL needs. Certainly the prospect
of DL students concentrated our minds and provided the stimulus. However, it
is noticeable that in some respects there is a danger of over-teaching; [residential]
students are expecting too much. Since so much more material is provided in
full printed form many students are now expecting everything to be provided.
For example, the simple weekly task sheet might be designed to encourage
students to read through a particular section more carefully by asking Verify
result 2.3 (which perhaps was given with It is easy to show that ..). Some
students are now expecting and waiting for full printed solutions. Of course,
this expectation can be lessened by more careful explanation in the routemap
and study guide, though clearly some students will nevertheless resent this
24
apparent omission in provision. Ultimately, the only solution is to refuse firmly
and resolutely.
Some Problems Arising
Apart from the various problems and difficulties indicated above, many of us
found problems arose because of a feature not mentioned earlier. This is that
some of our lecture courses are given in common with undergraduates, at least
in part. For example, Medical Statistics is taken not only by most of the MSc
students but by also around 100 undergraduates from our level 3/4 MMath and
level 3 BSc programmes; Multivariate Data Analysis is taken by around 50
undergraduates. These students had no access to WebCT (we did not want the
discussion boards to become unmanageable) so they had no access to the web
material, in particular to the various data sets and outline solutions which are
easily handled by WebCT. Full lecture notes could of course be printed in
advance and distributed. An easy, though time consuming, solution was to
provide what were essentially duplicate web pages for this material that all
undergraduates could access, (the URL for some of these is given below). It is
a drawback that there seems no easy way of transferring material loaded into
WebCT directly to ordinary web pages.
A more pressing problem, arising particularly because of the presence of large
classes of undergraduates is What Are Lectures For? If all the web material is
self-contained and all solutions etc are made available on the web, what can
be done in the lectures that is useful, maintains interest yet is not essential for
the understanding of the course (otherwise it would have to be provided on the
web for the DL students and the circle bites back). This problem arises only
because of the presence of both residential MSc students (who are encouraged
to mix with the DL ones) and undergraduates. Consequently, it is not realistic
to hide the web based material from those attending the lectures. Attempting
to read out the lecture notes which the students have in front of them is of
course a disaster. Even displaying the notes and talking around them, highlighting
key points, is unsatisfactory. One solution is to use the lectures to provide
more examples worked on in detail in the lectures (e.g. on the board or OHP),
more real time demonstrations of practical analyses in software packages (which
requires an LCD projector and laptop) and summary lecture notes displayed on
either OHP or by PowerPoint. Some of these can be provided to the DL
students at a later date and indeed one course did provide copies of the PowerPoint
slides on the web pages at the end of the course. The provision of weekly
news summaries on WebCT for the benefit of the DL students also encouraged
some of us to repeat this verbally in the lectures, not just to fill in time but
because some students are more distant than others. Similarly, some matters
arising in the discussion boards could usefully be repeated verbally in lectures
25
though there is some conflict in this since the residential students would find
that their verbal queries were initially only answered on the discussion lists but
then hear the answer repeated a few days later in a lecture. Happily, this did
not happen too often.
Next Steps
This past year the modules offered in DL format have largely been self-contained
individual courses. One exception was the Statistical Laboratory module which
has a minor element of collaborative work early in the course but a major final
project which is worked on collaboratively and presented collaboratively. It
was possible to defer this presentation to the residential week in June, just after
the examinations. We had intended, and had encouraged, that much of the
work on it would be done beforehand with DL and residential students
corresponding by email and exchanging PowerPoint files. Indeed we set up
some special WebCT project groups which would facilitate this. However, in
retrospect understandably, almost no work was done on this project until after
the examinations on day 2 of the June residential week, after the DL students
had arrived.
Next year we shall be providing the Data Analysis module in DL format. This
module currently requires much collaborative work and collaborative report
writing and presentations, sometimes with rle play. These are part of the
professional skills aspects of the course. We are now beginning to consider
seriously how we can achieve our aims for this course but can as yet offer no
satisfactory solution other than saying that You must do it and suggest that
joint presentations can be produced in PowerPoint which can be exchanged by
email just as easily as Word files for the reports. The rle play elements could
be deferred until a residential week but this would alter our current schedule
where we aim to complete the sequence of submitted reports well before the
end the second semester. A point emphasised by the DL students in the feedback
meetings during residential week was their need for some flexibility in work
patterns and that this might conflict with wishes of the residential students. Of
course, this can be regarded as an added feature that our students will benefit
from coping with, both residents and DLs.
One future intention is to make more use of some of the facilities in WebCT for
submitting work and monitoring progress. The intention is that this will be
implemented for all the MSc students, though probably not for the undergraduates
at present. Indeed, currently resources do not allow for compulsory submission
of all exercises with detailed marking and feedback, certainly not for the larger
classes.
26
Some Technicalities
Software used in Sheffield for teaching statistics has largely been based on
Minitab at undergraduate level and with S-PLUS, WinBUGS and some SAS at
MSc level with occasional use of SPSS, though we are now moving towards S-
PLUS at undergraduate level as well (next year S-PLUS will be used exclusively
in first year). Our CHEST agreement permits students to buy copies of most of
these packages (on a one-year licence, renewable only with a new licence
number) for home use within the UK. We have thus had little difficulty in
software provision for the DL students. Additionally, we have required them to
use Microsoft Office which we cannot provide and so they have to purchase
this themselves. This was a reluctant decision and we did consider alternatives
but we felt it essential to avoid extraneous problems that could arise if we tried
to accommodate a variety of different systems.
We advise on minimum computing requirements but we cannot help with
purchase of hardware and so the DL students have to provide this themselves.
Some students are in employment where these are provided but the licence
agreements for software under CHEST typically conflict with installation on
machines to which other people have access.
Our WebCT pages are only accessible by registered users. Some of the course
material can be found on one of our personal pages where the undergraduate
users access it. This is at URL http://www.shef.ac.uk/nickfieller/ but we do not
guarantee that it will remain there indefinitely (solutions to exercises will be
hidden before the next cohorts of students arrive). This page also gives access
to the copies of the slides used in this presentation at the OU in May.
Details of our MSc course can be found at
http://www.shef.ac.uk/~pas/MSc
Illustrations of WebCT
We give below some illustrations of WebCT so as to give a flavour of the
environment, annotated in places. People who are familiar with the package
will see that our use is rather non-standard since we use it as one course for the
whole MSc. This affects the way we have provided the material for individual
modules.
Figure1 shows the opening welcome page and the signal that material has
changed since it was last accessed. Figure 2 gives the home page with indica-
tors of access to the individual modules. Since PAS6011 (Multivariate Data
27
Analysis) is not currently being taken by 3-year DL students they do not have
access to it. Pages for individual modules may be linked contents pages as in
or can just be similar to the home page. A disadvantage of the linked contents
page is that it is not possible to date-release individual chapters in sequence,
however moving through from one chapter to another is facilitated.
Figure 2
Figure 1
28
Figure 3 shows the index of discussions and the presence of new discussions on
one modules board. It is a quick job to check for the presence of new discus-
sions from time to time and whether they need to be dealt with urgently or not.
WebCT also has an email facility which can be automatically redirected to the
home email to alert for more pressing queries.
Figure 3
Figure 4
Often, supposedly urgent queries raised by email are best left for more open
discussion though in some cases it is perhaps good to maintain goodwill and
preserve anonymity of the questioner, as in Figure 4.
The copies of the slides of this presentation referred to above give a few more
examples of WebCT for those who are interested.
29
Distance Teaching of Statistics to
Undergraduates at the Open University
Jane Williams, The Open University
At the beginning of this academic year, February 2002, there were 27,817
registrations on courses run by the Faculty of Mathematics and Computing.
This figure includes post- graduate students and students registered in Singapore.
Of these, 7,055 registrations were for level one entry courses and 1,863
registrations were for level two and level three statistics courses.
The ages of our students range from 18 to those who are over 65. As an example,
the following chart shows the age profile of students registered on M246 Elements
of Statistics for the February 2002 presentation of the course.
There are no entry qualifications and students are advised as to the best place
to begin their studies by Advisors and Academic Staff based in the thirteen
Regional Centres throughout the UK. To continue with the M246 example,
21% of the students were completely new to the University and 19% had
qualifications below those required for enrolment at a conventional university.
What courses do they study?
MU120 Open Mathematics is a 30-point entry course for students with only
basic (literacy and) numeracy skills. The course aims to build confidence in
mathematics and to incorporate mathematical thinking into everyday activities.
30
At the beginning of the course, students are introduced to statistical ideas
including summary statistics, index numbers, and graphical representations.
The use of a graphics calculator is integral and ideas of regression are introduced
towards the end of the course.
MST121 Using Mathematics is a 30-point entry course for students who are
more confident in their mathematical abilities. It is arranged in four blocks,
the final one being devoted entirely to statistics and covers basic rules of
probability, probability distributions including geometric and normal distributions,
sampling distributions, Central Limit Theorem, confidence intervals, hypothesis
testing and fitting lines to data. Use of computer software (Mathcad and OUStat)
is part of the course. Also presented in Singapore.
MDST242 Statistics in Society is a 30-point course at second level, suitable for
students from across the University who need to include some statistics in their
degree profile. Only basic mathematical skills are required. The course uses
statistical techniques to investigate everyday situations and to give an
understanding of statistical ideas - exploratory data analysis, basic statistical
inference and relevance of statistics to decision making. It does this through
investigating questions such as the following:
Are we getting better off?
Does class size affect pupils performance?
How much do pupils truant?
Is my child developing normally?
M246 Elements of Statistics is a 30-point course at second level. Ideally students
should have completed MST121 before registering for this course. It is an
introductory statistics course that emphasises the practical nature of the subject.
SSC (Student Statistical Calculator - non-standard software) is used to analyse
data and to develop understanding of statistical concepts such as exploratory
data analysis, estimating confidence intervals, hypothesis testing and regression.
Also presented in Singapore.
M343, Applications of Probability is a 30-point course at level 3. For this
course students need to be competent in algebra and calculus. The course
introduces models to describe patterns of events that occur in time and space,
eg earthquakes and occurrence of plant species, and situations that occur at
discrete time points, such as gamblers ruin. Probability models are developed
for events which may occur at any time, eg the spread of an epidemic. Other
situations involving probability include genetics and fluctuation in the stock
market. Also presented in Singapore.
31
M346 Linear Statistical Modelling is a 30-point course at level 3. Students
need a good understanding of the basic ideas of statistics before starting this
course, together with an ability to apply ideas and interpret results. It covers
statistical modelling where a response variable depends on one or several
exploratory variables and uses problems and data to explore questions, such as:
How well does a patient respond to treatment, given age and severity of
disease?
How do different strains of wheat compare when grown in various
conditions?
The statistical tool used is GENSTAT. Also presented in Singapore.
What awards do these courses contribute to?
The awards listed below are those that students studying a statistics course may
be aiming towards. Not all students are interested in an award. Students may
study MU120, MST121 and not do any further statistics. These students may go
on to take awards not listed here.
Diploma in Statistics (120 points)
M246, M343, M346 + 30 points from a list that includes MDST242
To obtain an honours degree, a student must study 360 points, including 120
points at level three.
BA/BSc, BA/BSc (Hons)
This degree can be tailored to a students own requirements - a student can take
courses from any area of the curriculum.
BA/BSc (Hons) Mathematical Sciences
BA/BSc (Hons) Computing and Mathematical Sciences
M246, M343 and M346 are optional courses for above two degrees.
BA/BSc (Hons) Economics and Mathematical Sciences
MST121, M246 and M346 are compulsory courses for this degree. M343 is
optional.
BA/BSc (Hons) Health Studies
BSc (Hons) Natural Sciences
BA/BSc (Hons) Social Sciences
MDST242 is an optional course for above three degrees.
32
How do we support our students?
Academic staff based in regional centres
Advisors and Student Support Staff based in regional centres
Local tutors
Face to face tutorial sessions and/or dayschools (optional)
Phone and email contact with tutors and regional centres
Correspondence tuition - this is the major part of a tutors job; quick
turnaround and positive encouragement is vital.
Telephone helpline - operated by academic staff from the Faculty
Computer conferences - run by OUSA (Open University Student
Association)
What is next for the Statistics department?
BM240 Quantitative Methods in Business is a 30-point course at second level,
written specifically for the undergraduate business studies degree and is due for
first presentation starting in February 2003. The aim of the course is to help
students make sense of the large amounts of quantitative data produced by and
for business. Students will learn to perform simple quantitative analyses using
EXCEL and to interpret results of analyses performed by others. Students require
basic mathematical skills.
M248 Analysing Data is a 30-point course at second level, which will replace
M246, using MINITAB to replace SSC. It is due for first presentation in February
2003.
U247, Statistics in Society is a 15-point service course to be presented in place
of MDST242. First presentation is scheduled for February 2005. The course
will use SPSS which is particularly suitable for students from the Social Science
Faculty and from the School of Health and Social Welfare.
33
Java Applications for Teaching Statistics
Andrew Bertie, The Open University
M248 Analysing Data is a new introductory statistics course presented by the
Open University from 2003. As the course title suggests, the emphasis is on
learning statistics through analysing data rather than mathematics. MINITAB
will be provided for data analysis, but it was decided to develop further software
to support the teaching of more difficult statistical concepts. In the past the
OU has developed such software in the C++ programming language. For this
course, however, it was decided to develop the software in Java.
Advantages of Java
You may have already come across Java programs in the form of applets
embedded in web pages. There are already hundreds of statistical applets, and
a few complete courses in statistics that use applets, available on the web. But
Java can be used to write stand-alone applications just as in C++. The M248
Java applications and Suns Java Runtime Environment (JRE) will be delivered
to students on the same CD-ROM as MINITAB and the course data files, saving
students the costs of downloading the JRE and accessing the applets over the
Web problems highlighted by Hunt and Tyrrell (2000). So why choose Java
and not C++? There are a number of general advantages in choosing Java as a
development language that flow from it being designed for the Web.
The most important of these is its platform independence summed up by Suns
slogan Write Once, Run Anywhere (WORA). A program written in Java should
require no modification to run on PCs (with their different versions of Windows),
Macs, Unix or LINUX systems. This is an important advantage in distance
teaching, where students can own a wide range of PCs with different
specifications, ages, and operating systems.
There are some features of Java that make it particularly attractive to developers
of statistical software. Its platform independence extends to the storage of
numerical data types, ensuring statistical algorithm produces exactly the same
numerical accuracy and behaviour on any hardware or system.
A striking feature of Java is the platform independence of its graphics. The
modern PC provides exciting opportunities for interactive, dynamic and animated
statistical graphics particularly important in teaching statistics. But, even
34
more than programming languages, computer graphics has suffered from lack
of portability and standardisation. Traditional programming languages, such as
Fortran and C, lacked standard graphical libraries that would allow the
statistician to build plots from graphical primitives. The Java standard libraries
provide superb platform independent support for graphics. It has all the graphical
primitives you would expect, plus sophisticated handling of colour, stroke, text,
geometry, transformations, animation and rendering. These libraries make it
possible to construct any kind of statistical display, from basic exploratory plots
to stunning professional-looking presentation graphics. They also make it
possible to develop animations for educational use. Unlike other languages,
any graphical code you write using these libraries will run on any system capable
of running Java - in keeping with Javas WORA philosophy.
The Java standard libraries also provide excellent platform independent support
for building graphical user interfaces (GUIs). Using these libraries you can
design a modern graphical interface to your statistical application involving
multiple windows, dialogue boxes, menus, buttons, icons, fields etc. Remarkably,
Javas WORA principle applies here too: the code you write will work on different
platforms and operating systems without modification. Java allows you to choose
a look-and-feel for your interface. You can choose a look-and-feel to suit
Windows, Unix or Macintosh computer systems, or you can specify a standard
Java look-and-feel that will look the same on all these platforms. No other
programming language or system offers this power to design platform independent
graphical user interfaces.
The World Wide Web is by definition an international network, so Java is designed
to cope with differences between countries and languages. It provides support
for internationalization - the process of enabling a program to run correctly in
any country. With increasing globalization of distance education, this is an
important advantage. A minimal level of internationalization of data display
is possible with little effort. It is easy to write a Java program that automatically
detects the geographical location it is running in, and formats numbers,
percentages, currency, dates and times appropriately.
There is increasing pressure on universities to make their software accessible to
disabled people. Javas accessibility package supports assistive technologies -
such as audible text readers and screen magnification.
Java is virtually free. You can download everything you need to develop,
compile and run Java programs for the cost of an Internet connection. The Java
Development Kit (JDK), various utilities, tools, documentation and tutorials are
all free from Suns website, and Sun permit the free redistribution of the JRE.
35
All you need to write Java code is a basic text editor, although specialized Java
editors are also available for little or no cost.
What is Java?
Java is a structured language descended from Algol (through C), but the most
important thing about it is that its object-orientated (O-O). Essentially this
means that the language supports designing a program that closely models the
entities/objects that you are interested in, and the interactions between them.
These entities can be abstract concepts, such as a statistical distribution. This
approach is quite different to a traditional FORTRAN program say, that is
essentially a collection of functions/algorithms. O-O ways of thinking are
essential in the case of GUIs.
As an O-O language, Java is designed to overcome the problems of traditional
languages such as FORTRAN and C, and encourage the development of correct,
readable, easily modified and reusable programming code. Javas O-O approach
can be applied to statistical concepts, methods and data in powerful and
expressive ways that would be difficult in traditional languages. All this could
also be said of C++, but Java is simpler, easier to learn and less prone to
frustrating run-time errors than C++.
The Java language itself is quite small and quickly learnt the syntax and
basic data types are very similar to C. Beyond that is the class that represents
an object, similar to C++. All data (variables and constants) and methods
(functions) must belong to a class (unlike C++). Classes can inherit from other
classes, so you can have an hierarchy of classes. The real skill in Java
programming is choosing classes that produce an elegant and general solution
to a problem, which can be easily reused and extended.
Sun provides a vast number of ready-made classes known as the Java API,
which are fully documented on Suns website:
http://java.sun.com/products/jdk/1.4/docs/api
There are also class libraries provided by other vendors, some free, some for a
price. For example, a good free library for matrix computations is documented
at http://math.nist.gov/javanumerics/jama/doc. The Sun API classes include
the usual mathematical functions, random number generators (including one
for the standard normal). There are classes for displaying and editing tables,
and the 2D Graphics API provides graphical primitives. Beyond that there are
no specifically statistical classes which led me to develop my own class
library called Java Statistical Classes (JSC).
36
For a more detailed introduction to Java, see Murdoch (1999). Chambers(2000)
describes Java as a more powerful engine for statistical software than, for
example, C or C++.
The advantages of implementing statistical systems in Java are discussed at
length by Lang (2000).
SUStats
SUStats (Software for Understanding Statistics) is a suite of 21 Java programs
(applets/applications) designed to support the teaching of some statistical
concepts in M248. They support the teaching of:
The relative frequency settling down concept of probability.
Sampling distributions.
Modelling variation and data distributions.
Central limit theorem
Confidence intervals
Hypothesis testing, power
Parameter estimation
Least squares regression
Figure 1 shows one of the SUStats applets that demonstrate the relative frequency
concept of probability. The Rolling a die applet simulates an experiment
involving rolling a die in which two faces of the die are defined as success
and shows how the relative frequency of successes settle-down to values close
to 1/3. The die shown below the graph is rapidly spinning, but not randomly; it
is cycling through its faces from one to six in sequence. The student clicks on
the Stop button to stop it spinning and select the face value shown on the die,
but the die is spinning so rapidly that their selection will in effect be random -
Figure 1: The Rolling a die applet
37
like rolling the die. This follows a suggestion of Velleman and Moore (1996)
that software should allow students to make random selections, since they may
not regard computer-generated random numbers as genuinely random. As
selections are made, the relative number of successes are plotted in the graph
and the current proportion is displayed numerically.
Figure 2 shows the Hypothesis testing applet that simulates fixed-level and
significance testing of the means of six standard distributions and an arbitrary
distribution called Your distribution that the student defines by manipulating
its pdf curve. There is a tabbed panel for each of the distributions. Figure 2
shows the applet with the Your distribution tab selected. Each panel has the
same layout. The population is represented in the top-left of each panel by a
plot of the distribution and fields for its parameters, corresponding mean and
standard deviation of the population. In the bottom-left area of each panel
are controls for the simulated testing, consisting of fields for specifying the test
mean, alternative hypothesis, significance level, sample size and number of
samples, a check box for switching to single-step mode, and buttons for starting
the simulation and resetting the applet.
The right half of the applet, consisting of two tabbed panels labelled Fixed-
level tests and Significance tests, displays the simulation results. When the
Take samples button is clicked, the specified number of samples of the specified
size are randomly generated from the population. For each sample, the test
statistic and significance probability are calculated. In the Fixed-level tests
panel the value of the test statistic is counted if it lies in the rejection region
and summary results about the simulation appear below the plot of the null
distribution plot. In the Significance tests panel the values of the significance
probabilities are plotted in a histogram below the null distribution plot.
Figure 2: The Hypothesis testing applet
38
Figure 3 shows the Principle of least squares applet. Students can enter data
points by clicking on the plot, typing or pasting data into the table, selecting a
real data example or generating a random sample. The scatterplot in Figure 3
shows a fitted least squares line, confidence intervals of the mean and prediction
intervals. The overlapping squares represent the squared residuals. The influence
of individual data points can be explored by dragging their markers with the
mouse or editing their values in the table the fitted line, intervals and squares
changing immediately. An option is provided to fit the least squares line through
the origin. Clicking on the Yours button allows the student to fit a line by hand
manipulating its position using three beads.
Open University students will use the stand-alone application versions of these
applets with printed computing activities, but basic instructions and suggested
activities can be found in the web pages containing the applets at
http://www2.open.ac.uk/CES/projects/SUStats/SUStatsApplets.html
The JSC library of reusable classes that supported these developments is being
extended to cover a wide range of statistical methods and graphics and will be
made available to the statistical computing community when completed.
Acknowledgements
The development of SUStats was a collaborative effort of the M248 course
team consisting of members of the Open University Statistics Department. Gillian
Iossif (course chair), Paul Garthwaite, Paddy Farrington, Pia Larsen, and Kevin
McConway provided ideas for applets and contributed to their design .
Figure 3: The Principle of least squares applet
39
References
[1] J M Chambers (2000) Users, programmers, and statistical software,
Journal of Computational and Graphical Statistics, 9, 404-422.
[2] Hunt, N and Tyrrell, S (2000). Learning statistics on the web
DISCUSS.Teaching Statistics, 22(3), 85-90.
[3] Lang, D T (2000). The Omegahat environment: new possibilities for
statistical computing. Journal of Computational and Graphical Statistics,
9, 423-451.
[4] Murdoch, D J (1999). Whats Java? Chance, 12, 44-47.
[5] Velleman, P Fand Moore, D S (1996) Multimedia for teaching statistics
promises and pitfalls, American Statistician, 50, 217-225.