penny bidgood, kingston university, uk assessment matters – original assessment for original...
TRANSCRIPT
Penny Bidgood, Kingston University, UK
Assessment Matters – original assessment for original student work, HEA York 2011
Why assess?
Who is being assessed/ assessing?
What should be assessed?
Where is the assessment taking place?
How is the assessment done?
HEA, York, 2011
Assessment –processes that appraise knowledge, understanding, abilities or skills Promoting learning by providing feedback Evaluating knowledge, understanding, abilities, skills Providing a grade Enabling public and HE providers to know attainment
level
“Diversity of assessment practice between and within different subjects is to be expected and welcomed”(http://www.qaa.ac.uk/academicinfrastructure/codeOfPractice/)
HEA, York, 2011
In implementing assessment policies consult subject bench marks and professionals
Types of assessment should be appropriate for subject, mode of learning and the student
Promote effective learning
Effective and appropriate measurement but avoid excessive burden for students
Provide appropriate and timely feedback
HEA, York, 2011
Specialist statistics or Service modules
Similarity: analyse data appropriately and report results effectively (exploratory data analysis / statistical modelling)
Differences: class size depth of mathematics computer package used
HEA, York, 2011
Students have different strengths andapproaches to learning and may perform differently in various types of assessment
Student Voice
A variety of assessment tools are required
HEA, York, 2011
Learning outcomes:-“Communicate technical ideas in writing” Lecturers:- workloadsResearch takes priorityRe-assessment issues
Design different types of assessment
HEA, York, 2011
Four Themes:-
1.Relating assessment to real world problems
2.Assessing statistical thinking3.Individualised assessment methods4.Assessing problem solving
HEA, York, 2011
Statistics as “mathematics” Formal examinations and tests
Statistics as an “applied subject” Assignments and projects
Statistics in a consultancy Oral and written reports portfolios
HEA, York, 2011
Understand the purpose and logic of statistical
investigations the process of statistical investigations mathematical relationships probability and chance
Master procedural skills
HEA, York, 2011
Develop interpretive skills and statistical literacy ability to communicate statistically useful statistical dispositions
“The Assessment Challenge in Statistics Education” (1997) Gal I and Garfield J B
“Assessment Methods in Statistical Education: An International Perspective” (2010) Bidgood P, Hunt N & Jolliffe F (eds)
HEA, York, 2011
1. Emphasise statistical literacy and develop statistical thinking;
2. Use real data;3. Stress conceptual understanding rather than
mere knowledge of procedures;4. Foster active learning in the classroom;5. Use technology for developing conceptual
understanding and analyzing data;6. Use assessments to improve and evaluate
student learning.
HEA, York, 2011
Garfield (1994, 1995) stressed the need for assessments that measure the understanding of a problem solving approach
The Mathematics, Statistics and OR Overview Report (2000) stated “Student engagement and performance has often been greatest when dealing with well-focused problems of a practical nature”
HEA, York, 2011
Using real, or at least realistic, datasets in an appropriate context of real problems.
Led by reforms in statistical education that emphasises statistical thinking, reasoning and conceptual understanding
Demands from employers – graduates with technical skills and the ability to communicate findings appropriately
HEA, York, 2011
HEA, York, 2011
English National Curriculum
HEA, York, 2011
Computer lab◦ On-line testing from a large databank◦ Using a computer package to analyse data◦ Practical assessment
Formal examination
In the lecture room ◦ Often difficult to supervise adequately
In a seminar
HEA, York, 2011
Examination Quizzes Multiple choice test (question bank) “Take-home” assignments Coursework – apply specific techniques to particular problem Analyse computer output Group work assessment Design and carry out a statistical investigation Prepare a report and present it Oral presentations Case Studies/Projects
HEA, York, 2011
One week before students are given a numerical question with a small amount of data – just to illustrate the data and their layout.
In the examination proper students are given the full set of data and access to SPSS.
They are required to explain their choice of test, briefly describe the SPSS commands, report summary statistics and draw appropriate conclusions.
C. Dracup, Northumbria (PiSA project)
HEA, York, 2011
Which of the following statements about the application of one-way ANOVA is false ?
The null hypothesis being tested is that all of the underlying means for all groups are identical
The F-statistic compares the variability between the sample means with the variability within samples
Large values of the F-statistic provide evidence of a difference between the underlying true means
If the P-value is large, this implies that all the means are the same A small P-value suggests that the data provides evidence of
differences in the underlying mean between some of the groups
Assessment on a Budget Wild et al in Gal and Garfield
HEA, York, 2011
Issue a published paper for students to review in advance and answer question(s) on it in a formal exam setting
Give students case studies throughout the year, which they are free to discuss with each other, but the assessment on the case studies is in the form of a supervised test
HEA, York 2011
The standard statistics assignment◦ analyse of a set of data, using a suitable package, and
submit a written report
Plagiarism concerns can have a strong influence on assessment strategies
Group or individual assessment?
Move away from “take-home” assignments, particularly in large service modules.
Plagiarism in Statistics Assessment (PiSA) http://www.jiscpas.ac.uk/documents/pisa.pdf
HEA, York, 2011
The RSSCSE and the MSOR Network jointlyfunded the PiSA project, which aimed to:
survey HE lecturers in Statistics to find out what methods of assessment and strategies to deter plagiarism are being employed currently;
to identify and synthesise elements of good practice;
to disseminate findings widely.
HEA, York, 2011
Plagiarism is “To take and use as one’s own the thoughts, writings or inventions of another” (OED)
Collaboration is to work together for mutual benefit
Collusion is to work together for mutual benefit with the intention to deceive a third party
HEA, York 2011
Is coursework important in Statistics? Are Statistics lecturers alert to plagiarism? Is plagiarism causing a reduction in
coursework? How are Statistics lecturers tackling plagiarism? What good practice can we share? Not prevalence Not case history Deterrence is the key
HEA, York, 2011
Institutional proceduresOrganisational measuresSupervised assessments Individualised assessmentsStudent-centred assessmentsElectronic submission
HEA, York 2011
Plagiarism, or more specifically collusion, is a significant problem within large Statistics service modules and all lecturers need to give serious attention to anti-plagiarism assessment strategies.
The majority of Statistics lecturers are well aware of plagiarism issues and are taking action, however small, to combat it.
It is quite common for Statistics lecturers to fail to apply institutional procedures in “minor” cases of plagiarism. In contrast, some lecturers make every effort to demonstrate how the regulations and penalties might apply to Statistics assessments, giving examples of cases detected in previous years.
HEA, York, 2011
Plagiarism often goes undetected on large service modules due to a multiplicity of assessors. It is most likely to be detected when one person assesses all the students.
There is much innovative work taking place in the area of individualised assessment, but also some duplication of effort.
Assessments that require students to collect their own data, either individually or in small groups, are widely employed.
Many lecturers have moved away from take-home assignments to in-class supervised computer-based assessments.
HEA, York, 2011
In-class tests can be exposed to a high risk of cheating by unsuitable accommodation, inadequate invigilation, failure to check student identities, and naïve organisation.
TURNITIN is being used increasingly. In projects/case studies it is good practice to
include an element that assesses the student’s working method and, ideally, an oral to check that it is genuinely the student’s own work.
Online cheating companies openly offer an easily accessible way for students to obtain professional individual help with Statistics assignments.
HEA, York, 2011
Randomise elements or parameters
Allocate different subset of a large dataset e.g. ISCUS – Individualised Student Coursework
Using Spreadsheets. This is based on Excel and allows you to use your own dataset. (Developed by Neville Hunt, Coventry University)
Students allocated subset of same source data based on ID number ABCDEFG
o e.g. delete rows C, D and Eo or calculate a 9G% confidence interval
HEA, York, 2011
Students find own data, from journals,internet sources, about themselves
Allocate a particular periodical – data from any issue in current year
Medical statistics – find own example of a medical case study
Time series modules, find own data from “Economagic”
Sports science – collect data on fellow students e.g. heart rates
HEA, York, 2011
Report in form of poster, oral presentation, written as for a newspaper article etc
Vary format of the submission ◦Posters, typically produced by a group of
students, although each should be able to “defend” the content
◦Written report, but vary the “client” – newspaper article, research paper, briefing document for local MP etc
HEA,York, 2011
Assignment of sub tasks within a group Degree of input from each participant Quality of final product from group Individual learning which has taken place
Peer Assessment Some element of marks might be how they rate
others’ contribution and their own
HEA,York, 2011
Need to include an element that assesses the student’s working method
Ideally an oral exam or presentation “4P model”
◦Project log◦Project report◦Practical development◦Presentation
(Sue Starkings, reported in Gal and Garfield, 1997)
HEA, York, 2011
Aims To make available real datasets and scenarios of
relevance to Business, Health and Psychology
To develop web-accessible statistics worksheets using these datasets and various software (Excel, MINITAB, SPSS)
To develop resources for producing individualised datasets and assignments
Funded by the Higher Education Funding Council for England October 2002-January 2006
HEA, York, 2011
ISI, Durban, 2009
Chatfield (2005) – difficult to teach using a problem-solving approach Problem Solving: A Statistician’s Guide
Rossman and Chance (2002) developed materials, motivated with real data and scenarios, using various problem-solving skills. ICOTS Proceedings
Jolliffe (2007) “Asking students to do real statistics on real data and to report on the results, is now feasible in a way that it was not in the past” (due to huge expansion in technology) http://www.stat.auckland.ac.nz/~iase
Marriott et al (2009) developed assessment regimes that correspond to the problem-solving approach in teaching and learning www.amstat.org/publications/jse/v9n3/
HEA, York 2011
http://www.qaa.ac.uk/academicinfrastructure/ http://stars.ac.uk http://app.gen.umn.edu/artist/ http://www.jiscpas.ac.uk/documents/pisa.pdf http://www.amstat.org/education/gaise/GAISEcollege.htm
HEA, York 2011
Bidgood P, Hunt N & Jolliffe F (eds) (2010) “Assessment Methods in Statistical Education: An International Perspective”
Gal I and Garfield J B (1997) “The Assessment Challenge in Statistics Education” o Starkings, S. Assessing Student Projectso Wild et al Assessment on a Budget Garfield J B (1994) “Beyond Testing and Grading: Using Assessment to improve
Student Learning” Journal of Statistics Education Garfield JB (1995) “How Students Learn Statistics” International Statistical Review
63 1 Holmes P (2004) “Assessment in Statistics: a two-edged sword” in Assessment
with a Purpose Conference Hunt D N “Individualized Statistics Coursework Using Spreadsheets” Teaching
Statistics 29 2
MSOR Overview Report (2000) Quality Assurance Agency for HE
HEA, York, 2011
40
41
42
Generators for individualised datasets, assignments and solutions
ISCUS – Individualised Student Coursework Using Spreadsheets. This is based on Excel and allows you to use your own dataset as well as those from STARS. (Developed by Neville Hunt, Coventry University)
DRUID – Dynamic Resources Using Interesting Data This is not tied to any statistics package but uses specific datasets. (Developed at the RSS Centre for Statistical Education, Plymouth University)
HEA, York, 2011
Q5. Why is the clustered column chart unsuitable for the age data?
We are now going to draw histograms of the ages for each of the treatment
groups so that we can compare them From the main menu select Graph > Histogram Highlight Simple and click on OK Enter Age in the Graph variables: box Under Multiple Graphs > Multiple Variables choose In separate panels of the same graph and under Multiple
Graphs > By Variables enter Treatment group in the By variables with groups in separate panels: box Click on OK Under Scale> Y-Scale Type choose Percent Click on OK Enter an appropriate title as before Click on OK Click on OK again to produce the chart below.
HEA, York, 2011
There are so many different ages they become cluttered rather than clustered! The ages need to be grouped into intervals.
Carlow, 2010
Age
Perc
ent
645648403224
30
25
20
15
10
5
0
645648403224
New drug Placebo
Distribution of patients' ages
Panel variable: Treatment Group
HEA, York, 2011
Q6. Is there an evident difference in the distribution of patients’ ages between the two groups?
Height and baseline weight
Try replacing C2 (Age) by C4 (Height) in the analysis above.Q7. Does the distribution of patients’ heights differ between groups?
Repeat the analysis using the baseline weights in C9.Q8. Does the distribution of patients’ weights differ between groups?
Both distributions appear quite similar although there is a wider spread of ages for the Placebo group. The most common age for the Placebo group is 40-44, whilst for the New drug group both 36 -40 and 52-56 have the same frequencies.
Height has an almost uniform distribution in the Placebo group, but a more uneven distribution in the New drug group.
Weight has a slightly negatively skewed distribution in the Placebo group, while the distribution in the New drug group is bi-modal.