Teaching statistic in the context of the research process: An inquiry based approach

1
Teaching statistic in the context of the research
process An inquiry based approach

• Richard Rowe
• Department of Psychology
• University of Sheffield

2
Inquiry Based Learning

• Learning through self-directed exploration
• Open-ended scenarios
• Different approaches may provide valid solutions
• Students choose the methods employed

3
Transferable inquiry skills

• Information literacy
• Collaboration
• Interpersonal skills
• Reflection
• Independent learning

4
Research Methods and IBL

• Statistical analysis is not inherently inquiry
  based
• Psychological research IS an inquiry task
• Psychology students not always enthusiastic about
  research methods
• Become interested during Level 3 project
• When using statistics as an inquiry tool

5
The quantitative research process
Hypothesis generation
Topic choice
Literature search
Experimental design
Analysis
Data Collection
Reporting
Interpreting
Write-up
6
Inquiry based revisions

• Combine with didactic lectures
• Want to cover the whole research process at
  Levels 1 and 2
• Introduce
• Information literacy
• Hypothesis formation and experimental design
• Collaboration
• Develop
• Analysis and interpretation in context of
  research literature
• Transferable skills
• interpersonal, communication, reflection,
  independence

7
Resources available

• Level One
• 4 postgraduate led tutorials
• 2 coursework assignments
• Level Two
• 2 lab classes
• 2 coursework assignments
• (130 students)

8
Level One
Literature search activity
Hypothesis generation
Topic choice
Literature search
Experimental design
Analysis
Data Collection
Reporting
Interpreting
Write-up
9
Research design tutorial

• Tutors present choice of research questions
• From (postgraduate) tutors field
• Collaboratively design study to test one
• Fill out summary sheet at end of tutorial
• Hypothesis, IV levels, Design choice, DV choice,
  confounds

10
Analysis Workshop tutorial

• Collaboratively analyse data from experimental
  design
• Data simulated to have properties expected from
  collected data
• More variables than specified to allow all Level
  One Tests

11
Analysis Workshop tutorial

• Analyze and write-up data in one hour session
  (groups of 4)
• Choose statistical tests, run tests, interpret,
  write-up
• 2 Laptop computers provided (running SPSS and
  Word), notes and books allowed, Tutor as a
  resource
• Submit group report at end of session

12
Analysis worksheet

• Describes variables in data set
• 5 questions address
• T-test (related)
• T-test (unrelated)
• Pearson correlation
• Chi-square
• Generate own question and test

13
Example data description

• age (measured in years), sex (0f 1m)
• driving aggression (dragg) pre intervention
  scored on a questionnaire, range 0-20, higher
  scores means more aggressive
• driving aggression post intervention (dragg2).
  The same driving aggression scale conducted at a
  later date. In the interim all drivers had taken
  part in an anger management course.
• Ever had an accident (accid 0no 1yes)

14
Example questions

• 1. Is age related to pre-intervention driving
  aggression?
• 2. Are male drivers more aggressive than female
  drivers (pre-intervention)?
• 3. Are male drivers more likely to have been
  involved in an accident?
• 4. Did the anger management program reduce
  driving aggression?
• 5. Choose one further analysis to run using your
  dataset.

15
Conceptual evaluation

• Test interpretation and reporting
• Also test
• choice of statistical test
• Use of SPSS
• Without risk of unfair collusion
• Ability to construct research questions
• Group obscures individual assessment
• 80 course assessed via exam
• Too many questions in assignment!

16
Tutor led assessment

• Shared responsibility for daunting task
• Tutor helps to ensure success
• Mimics availability of real-world expert resource
• Unsure how much help to give
• Tutorial is a learning environment
• Assessment provides motivation to engage
• Fewer scripts to mark
• Easier to provide detailed feedback

17
I feel more able to formulate research questions
69 Agree
18
Tutor feedback

• Design tutorial
• I have learned that by speaking in an
  enthusiastic way about an area of research, it is
  possible to inspire students to take a deeper
  level of interest.
• Analysis workshop
• The students seemed familiar enough with the
  material being tested, but had trouble applying
  their knowledge to a practical problem.
• The students completed the task within the set
  guidelines and seemed pleased with their
  work-although they may just have been relieved it
  was over!

19
Problems with collaboration

• Group work just tends to be left to one person
  who has to take responsibility for it.

20
Level 2Putting it all together
Hypothesis generation
Topic choice
Literature search
Experimental design
Analysis
Data Collection
Reporting
Interpreting
Write-up
21
Level 2

• Semester One
• Research proposal
• Analysis and report simulated data
• Focus on 1-way ANOVA
• Semester Two
• Full experimental design and report
• Focus on multiple regression

22
Semester 1 Research proposal

• Work in groups of 4 during lab class
• 2 computers per group
• Choose research topic between group
• Search for papers on topic using Web of knowledge
• Design experiment to follow-up
• Constrained to design suitable for 1-way ANOVA
• 1 variable, 3 levels
• Summarize proposal during class

23
Data generation and analysis

• Dataset simulated to mimic that which might have
  been collected
• Personalised versions for each student
• Analysis in own time
• 1 way ANOVA (all significant)
• Post hoc test (all 2 groups same, 1 different)

24
Semester 1 Submission

• Individual research proposal
• Comprises literature review
• Details hypothesis and proposed methods
• Individual results write-up
• Write-up analysis of simulated data
• 10 of module mark

25
Example proposals

• Using a matching pictures task to study telepathy
  in monozygotic twins, dyzygotic twins and
  siblings
• The effects of opiate antagonists on pathological
  gambling
• The effect of alcohol consumption on reaction
  times to potential driving hazards in young men

26
I feel more confident using Web of Knowledge
71 Agree
27
I feel more able to design experiments
78 Agree
28
I feel more able to apply 1-way ANOVA in SPSS
97 Agree
29
I am more confident about writing up statistical
analyses
96 Agree
30
I prefer analysing data based on my own design to
data on topics chosen by the lecturer
42 Agree
31
The project developed my skills as an independent
researcher
76 Agree
32
Evaluation

• Historically research methods has lots of maths
• Challenge to fill the void
• Conceptual understanding of statistical tests
• Full engagement with entire research process
• Inquiry based methods well suited
• Also develops transferable skills
• Further test of efficacy will be project students
  in 2008-2009

33
Acknowledgements

• Psychology Department
• Paul Norman
• Myles Jones

• CILASS
• Pamela McKinney
• Jamie Wood

Kamal Bhana
34
Teaching statistic in the context of the research
process An inquiry based approach

• Richard Rowe
• Department of Psychology
• University of Sheffield