Back to MEANS Homepage Participants
Professor Martin Bland (St George's Hospital Medical School),
Stephen Bremner (St George's Hospital Medical School), Peter
Collinson (Office of National Statistics), Ian Cox (University of
Cranfield), Professor Neville
Davies (Nottingham Trent University), Simon Dunkley (SPSS), Anne Hawkins (RSS Centre
for Statistical Education), Peter Holmes (MEANS
Project), Philip Jarvis (De Montfort University), Rhodri Jones (Dept
of the Environment), Michael Morris (Office of National
Statistics), Ian Schagen (NFER), Susan Starkings (South Bank
University)
Peter Holmes and Anne Hawkins welcomed everyone and gave a
brief background to the MEANS Project and the work
that had been done so far. Peter described some of the findings
that had come through correspondence and visits; Anne reported on
the job survey.
The session opened with a free flowing discussion around the
topics of the workshop. This was followed by each participant
giving an individual presentation.
Mick Morris said that his first real statistics problem
at the Office of National Statistics had been to develop an index
for measuring the price of holidays. His undergraduate theory had
not really prepared him for the practical requirements of this
task (how to collect prices, which prices, who to contact, how to
set up the problem etc). Rhodri Jones found that he
quickly had to develop communication skills and write reports.
This was not part of his scientific or statistical background. Stephen
Bremner had had a better grounding in these things from
Management Science. He had not been taught much on the
presentation of statistical data. Neville Davies said that
these things were part of the courses at Nottingham Trent
University, but they did find a problem with the students' poor
quality of English.
Whilst interviewing for new statisticians Ian Schagen
noticed that applicants had difficulty in addressing real
problems even when details had been previously circulated. There
did seem to be a mismatch between theory at university and actual
practical problems. Communication skills should also include
reading skills.
Some of the statisticians working with Peter Collinson
acted mostly in a consultancy role; these skills were generally
not found in new recruits so he is organising courses on
consultancy skills aimed at higher management. These courses are
being run by Doug Zahn and Dan Boroto of Florida State University.
At South Bank University Susan Starkings does have an option
on consulting but students do not choose to do it.
There was a general discussion on learning during which the
following points were made. One person was taught how to learn in
a first year psychology course but he thought this too early.
Learning theory should be part of all courses and used in the
construction of courses. Different types of people learn in
different ways. A good course on learning helps you identify your
own weaknesses. Students can easily pick up bad habits. Some
thought it best to build interpersonal skills into beginning
courses and continue throughout. Students from other disciplines
could be used to check the ability to communicate. There is more
incentive to get this right when faced with a paying customer.
Universities may need to break down the insularity of students/courses
and encourage a mix of disciplines.
It might be useful to look at how Business Schools teach
presentation skills. There is a problem that even where some
courses require communication skills these are not proactively
taught.
Martin Bland suggested that we need to keep things in
perspective. At the end of a five year medical course we do not
expect graduates to do a job straight away in medicine. It needs
a follow-up year and postgraduate training. People need to learn
at work. Maybe something similar is reasonable for statistics
graduates. There is a role for professional development and the
professional societies.
Reference was made to a perceived attitude change in students
- more seem to be wanting a qualification than to be really
interested in learning a subject.
Philip Jarvis had noticed that a sandwich degree sorts
students out. After the year out they are more mature and more
realistic. On a separate issue he pointed to the wide scope for
opting out in modular schemes.
If done well the placement year is more than just a time shift.
The tutor has an important role in making it part of the learning
experience. There is also a spin-off into third year courses
where problems from the placement year can be used. It is in the
placement year that many students learn to work under pressure to
time restrictions and have to prioritise their work.
Simon Dunkley had an undergraduate background in
Psychology where the approach was 'here are the statistics you
will need for your practical'. His examinations were open-book.
He is not sure now that he understood the background theory or
why, for example, the normal distribution is useful. He worked as
a technician in a consulting role in a medical school advising
psychiatrists how to get their papers published. Currently at
SPSS as a trainer he continues to come across good visual
presentations of important statistical ideas. [He gave the
example of 3 normal distributions to illustrate 'between' and 'within'
variability for ANOVA]. SPSS does employ statisticians - who
sometimes have difficulty in coming up with simple explanations.
They improve as they become established.
Neville Davies described the courses he is teaching
this year. 190 students on a basic statistics course for 1st
year economists. A second year course on 'Communicating
Statistics'. A 3rd year course - dynamic graphics and
modelling. A third year course on forecasting for business. He is
using the Web to manage his courses. He collected a dataset from
his first year students (sex, coins in pocket, estimate of ND's
age, will FT index rise or fall etc) and puts it on a Minitab
Worksheet. He uses these data to illustrate the course. He has
tried putting his lectures on the Web (on Powerpoint slides) but
finds that then some students do not come to the lectures. Each
lecture has a brief evaluation from some students (what found
hardest, easiest, liked most, like least). The notes for the 'Communicating
Statistics' course are on
the web. Although designed for 2nd year honours
statisticians it was chosen as a module by 40/50 Business
students so he had to modify it.
He had one student critically evaluate the STEPS modules.
The student commented that he didn't realise he was learning.
Another had to search the Web for electronic statistics courses
and then evaluate. Another was asked to reflect on a first year
topic that she did not like then present it. He spends 2 weeks on
problem solving and uses role play. He includes practice at
statistical interviews.
Stephen Bremner has just started as a Research
Assistant at Department of Public Health Sciences working on air
pollution and mortality. He acts as a tutor for first year
medical students on research and critical skills and also works
with health professionals who may not have a first degree.
His course at Strathclyde University included included
mathematics; management science and computer science. Statistics
was taught from examples and used computers. This did motivate
him, and he did have to mix with other students. Minitab was used,
but only once per fortnight, so it did not really get established.
There were more statistics options in his third and fourth year.
More students chose the practically oriented courses than chose
the core mathematical theory. There was some group work and
continuous assessment carried more weight than examinations. Most
of the practical experience came from Management Science. He
spent a summer on a placement scheme with a Public Health
Authority and looked at asthma in a fairly unguided way. He had
better collaborative experience in his MSc. There were groups for
course practicals, discussion papers. He had nothing in his
undergraduate course on data management or using large data sets.
[Others commented. One found an MSc was needed to refresh and
diversify. One employer was only looking at MSc's and relevant
experience. Real Project work has resource implications. There
are practical problems in cleaning up a database. This process
can be used to teach elementary statistics. Many employees in
statistics have to be data managers.]
Martin Bland described his background from a degree in
Mathematics, Statistics and Operations Research through work with
ICI on Agricultural Chemical Research to medicine at St Thomas'
medical school and lecturing in epidemiology and teaching
statistics at St George's hospital. He employs statisticians to
analyse large data sets from epidemiological and health services
research. For example there are surveys of respiratory diseases
related to social and environmental factors; deaths due to asthma.
He uses his research staff for small group teaching. He has
stopped giving lectures. Notes are given, exercises are given to
expound on research. The aim is to teach students to read medical
literature. Final year medical students do a survey on a topic of
their choosing. They have to process and present it in a week. He
recruits 1 or 2 statisticians each year (in medical statistics)
they have to learn on the job (including SAS). When interviewing
he looks for the ability to work independently, for suitable
technical qualifications with no general feature missing. The MSc
courses in medical statistics are a good basis for what he wants.
An interesting survey appeared to show that as people move
from clinical medical students through junior house-doctors to
consultants their view of the value of statistics increases.
Philip Jarvis De Montfort University are trying to
tackle the difficulty of training for the real world. Their
courses are geared to fit in to real use and are aimed at a
particular market e.g. Medical and Health Statistics, Industrial
Statistics. The context is to get what companies want and use
industry to validate. Teaching uses seminars and tutorials,
brings in people from the industry and links with consulting.
Local companies are enthusiastic. He has used data and problems
from Zeneca, Knoll, Smith Kline Beecham. There is some difficulty
in recruiting for the courses and the module mix can be a problem.
Ian Cox From a physics background at Teeside
Polytechnic, Ian worked for 7 years in semi-conductor
manufacturing and then had to oversee the statistical aspects of
manufacturing operations. He employed some statisticians in a hi-tech
company. Now at Cranfield he is involved in service teaching the
statistical aspects of process management.
After the meeting Ian wrote to say that the paper by Boroto
and Zahn ('Promoting Statistics: On Becoming Valued and Utilized',
The American Statistician, 43, 71, 1989) come close to
encapsulating some of the experiences I have had, and tries to
give a historical explanation of how this state of affairs arose,
whilst offering some recommendations for action. At any rate, I
see the lack of systemic and statistical thinking in
Manufacturing as a limiting factor to our competitiveness, and
would assume that, in the medium to long term, the resolution of
these issues must follow from the recommendations of the MEANS
project and others like it.
Rhodri Jones is now working at the Department of the
Environment. He explained that the CSSB interview procedure had
asked fairly basic questions to test his statistical ability but
had placed a lot of emphasis on testing his personal and
communication skills. He is now working on a survey of housing
conditions - a population of 25,000 properties. The survey uses a
complex stratified sample, the theory of which was not touched on
in his degree.
After his degree in Cardiff he became a maths teacher. If he
presented his work in a boring fashion, the children would soon
ask "what's the point of this work?". He tried to adapt
his lessons so that they would emphasise an end-use to the work.
Maybe university lecturers need to do the same.
The statistical element of his work at the Government
Statistical Service is mainly on surveys, and this had not been a
major part of his mathematics/statistics course. His degree was
titled 'Statistics with Management Science Techniques' and the
practical management science side was more enjoyed by the
students because it could be seen as useful. This teaching was in
context; others were dry and theoretical and perceived as useless.
No communication nor report writing skills were taught.
The demands of an employer can come as a shock to
postgraduates. It would be useful if lecturers became more aware
of the skills and practices used by employers of statisticians,
so that they can better prepare their students for entry into the
workplace.
Ian Schagen With a background in physics Ian spent 9
years with BP modelling oil reservoirs. An interest in
uncertainty led him to do a part time MSc at Birkbeck. He then
lectured at Loughborough University in computer studies,
statistics and operational research. There is no statistics
department there but there are statisticians in different
departments. He is now with NFER where they conduct independent
educational research. Work is project based, led by workers with
statisticians as part of the team. Work pressure means tight time
deadlines. There are 9 statisticians and recently they have been
advertising for statisticians. In advance of interviews they sent
a simple problem for applicants to discuss. They also asked
questions about using cross-tabulations, and about statistical
significance. At interview there were many irrelevant replies and
also a poor grasp of English. NFER have also noticed that
universities tend to concentrate on design of experiments rather
than survey analysis - is survey analysis down-graded in academic
teaching?
In applicants for statistician posts at NFER they are looking
for: a reasonable theoretical grounding; ability to work with
other people; ability to communicate; ability to convert a
problem to statistics, analyse and report; work accurately. They
also need to be able to work on more than one job at a time.
[Peter Collinson reported that Statistics Canada has a 6-week
course on surveys where participants work in teams, carry out a
survey and have to make it work for a real client.]
Michael Morris carried out a survey on public attitudes
for Hallamshire Hospital Psychiatric Ward as part of his degree
in Applied Statistics at Sheffield Hallam University. His
presentation of findings was video recorded as part of the
assessment. He found his course helped him to become a problem
solver rather than a statistician. Even with no theoretical
background (or one forgotten) he can go back to his notes or
books to refresh. He felt he was taught the learning process. The
case studies and practical sessions were useful, linking to
lectures or to a consulting problem and often using the computer
laboratory. His one-year placement was at the Treasury in the
macro-economic forecasting section testing the stability of
regression coefficients and he did a project on forecasting
inflation. Immediately after graduating he did some consulting
with SIA Ltd and learned to meet tight time constraints. After 18
months he went through CSSB selection for Fast Track
Statisticians. His first job was on Retail Price Index work, now
he is developing an indicator to measure holiday prices. His
university course did not fully equip him for this - the
practical problems of collecting price data and how to
operationalise the problems. It is possible to buy in expertise
from outside. The Government Statistical Service does give
training courses and offers access routes to improve.
[Simon Dunkley agreed with the points on trial management and
operationalising a problem. Could these be built in to university
courses? Could a university give students tight time deadlines.]
Peter Collinson His first degree in Pure Mathematics
from Newcastle included some statistics. This was followed by a
part time MSc at Birkbeck. He now works in the GSS Human
Resources Development Branch. In GSS there are 700 professional
statisticians; 5,600 in GSS support grades. The annual Civil
Service Selection Board (CSSB) competition recruits about 20
statisticians; from 400 applicants they select about 140 for
interview.
Training is given to CSSB recruits to fill in the gaps. Some
specific skills would not be expected in university courses - e.g.
the workings of government and parliament. There is not much
technical statistical training; most is on management and other
non-statistical skills. In 1998 the development scheme changes
introducing a framework of competencies needed by government
statisticians. Assessment will be made at recruitment followed by
a two-year probation. They are going to broaden recruitment to
numerate people with a wide background. In 1998 there will be a
new grade of non-fast-stream statisticians. They are developing
modular training in all aspects, for all in GSS, based on the
skills and competencies analysis. The move is towards doing
training when it is needed.
Susan Starkings studied Computing, OR and Mathematics
at the University of Kent. The course was largely theoretical
with no practical contexts but it helped her learn to think.
There was a contrast with here Masters degree course at Sheffield
Hallam University. Now lecturing at University of South Bank she
gives many service courses in statistics. The 1st year
course in Financial Statistical methods has over 300 members
which makes practical work difficult. The numbers on mathematics/statistics
degrees are falling. Report writing is covered in other courses
but there does not seem to be much carry over by students into
their statistical courses. Students tend not to like giving
presentations; they do not like questions with no known answer (e.g.
flood modelling for the Thames barrier - how high should it be).
Small groups work well in the final year - group marks are
awarded.
The meeting then split into two groups - the employers and the
medics - to discuss what need to be done by MEANS
Employers Find out what works in terms of assessment at interview.
Set up a semi-permanent forum for employers and academics to meet.
How can GSS share with academia; communicate and feedback.
GSS can put their list of competencies on the Web; differentiate between statisticians and research statisticians.
Encourage more accurate job descriptions.
Work on Continuing Professional Development needs is being done - Southampton and Greenwich have refresher modules.
Could university courses
Is it possible to ease students into presentation skills so that they do not avoid these courses
What is effective in teaching presentation skills and problem
solving?
Medics Medical MSc courses seem to do a good job. They are well focused, taught practically and relevant to the field.
Areas of concern are: work presentation, report writing, communicating to non-statisticians, content of data, data management. These need time and experience to mature. Maybe some content has to go to keep these in.
There is a resistance to removing anything. It is also difficult to assess these skills.
Get students to do their own research projects.
MEANS should try to
disseminate good ideas.