### Stats-Day-WKSP-non-sampling-errors

```Evaluate Statistically Based
Reports ( AS 3.12)
Workshop 1
Dru Rose (Westlake Girls High School, Ministry of Education Study Award)
Rachel Passmore (Endeavour Teacher Fellow)
Angela Hawkins (Rosehill College)
What does AS 3.12 cover?
i. Polls and Surveys


Non-sampling errors and survey concerns
Sampling error (margin of error, 95% confidence intervals,
testing claims made - Workshop A)
ii.
Experimental and Observational Studies
Why teach AS3.12 ?






Statistical Literacy is an essential life-skill to function
effectively in the information age (Wallman, 1993; Gal, 2002)
Broadens students’ horizons, taking statistical understanding
beyond the classroom into the real world (a motivational
aspect for students in the trial)
External standard
Only pre-requisite is AS2.9 (possibly just 1.10)
Links to other standards students may be taking (formal
inference AS3.10, experiments AS3.11, bi-variate data AS 3.9)
Suggested timeline for 2013
 Polls and surveys :
• Introduction to non-sampling errors and survey
concerns ≈ 6 periods
• margin of error, 95% confidence intervals, testing
claims ≈ 7 periods
 Experimental and observational studies
≈ 7 periods
Evaluating reports involving polls and surveys
How should we teach this ? (some approaches that seem to work)
Teacher modelling + group work + individual practice
 Literacy support:
1. Pre-reading strategies-look at the structure of media reports, skim
2.
3.
reading to “get the gist”, identify and discuss problem vocabulary
record answers on a writing frame
Post-reading strategies- write a 3 sentence summary, note 2 positive
aspects, note 2 concerns, organise into a short report
How should we teach this ?
 Begin with familiar contexts and gradually move to
unfamiliar (examples in resource pack- ask students to find some too)
 Use “stories” to illustrate the various non-sampling errors
examples of their own)
 Rephrase student comments from informal to formal
statistical language (Encourage students to use the correct terminology
in their reports - a step up from AS2.11)

Allow Time (e.g. one report per lesson – one or two non-sampling errors
at time until all covered – can keep non-sampling errors and survey concerns
going even when moving on to look at sampling error)
What are polls and surveys?
Do you have an internet capable mobile
device with you today?
% of New Zealanders carry an internet
capable mobile device, survey finds”
“
A possible student poll :
What type of driving licence do you hold ? (full, restricted, learner,
none)
 Use poll to explore and extend students’ existing statistical knowledge
base regarding: target population, sample, random selection, making an
inference
 extend students’ view of sampling from classroom activities to sampling of
the wider NZ population
Students need an idea of how key characteristics such as : gender, age,
ethnicity, regions of residence, income are distributed across NZ population
(A common student misconception : for a NZ survey on health the sample
should have equal numbers of smokers and non-smokers)
 Look at methodologies used by research companies eg DIGIPOLL,
COLMAR BRUNTON, HORIZON etc.(in power-point on political polls in
resource pack)
Sampling
Sampling Error
(random process)
Non-sampling
Errors
Selection bias
Non-response bias
Self selection
Question effects
Behavioural considerations
Interviewer effects
Survey-format effects
Transfer findings
C1, L2, S9
Interviewer Effects in Racial Questions
In 1968, one year after a major racial
disturbance in Detroit, a sample of black
“Do you personally feel that you trust
most white people, some white people
or none at all?”
• White interviewer:
• Black interviewer:
C1, L2, S10
The resource pack for this workshop has 4 media reports:
 Driving infringements –Digipoll (good methodology), question
concerns-illustrates structure of media reports
 Road tolls – Horizon on-line panel, potential selection bias, non-response
bias, question concerns, funding concerns-”Worry Questions” sheet and
writing frame
 Potential for DVT in office workers – a more difficult context but
relevant and interesting to students -funding concerns, question concerns,
selection concerns- a chance to discuss definitions
 NZ-US military –Digipoll (good methodology), a move to a more
difficult context- demonstrate “reading aloud”-how to make meaning from
context-need to discuss background re ANZUS, NZ non-nuclear stance
Experiments & Observational
Studies
What is an Experiment?
• Participants randomly
allocated to treatments
• Variable of interest is
measured and results
compared between
treatment groups
• Causal claim can be
Dru Rose, Angela Hawkins and Rachel Passmore
KEY TERMINOLOGY
• Response Variables – the variable measured as the outcome of an
experiment. Measure blood pressure to assess effect of new drug.
• Explanatory Variables – the variable that attempts to explain differences
in the response variable
• Treatment- one or a combination of explanatory variables assigned by the
researcher
• Confounding Variables – variables connected to the explanatory variable
that may be the actual cause of the differences in the response variable.
Dru Rose, Angela Hawkins and Rachel Passmore
How to design a good experiment
• Random Allocation – participants randomly allocated to
treatments and random order of treatment
• Use of a Control Group – use to establish base line effect
• Use of a Placebo – dummy, existing or no treatment
• Use of Blinding – single and double blinding
Dru Rose, Angela Hawkins and Rachel Passmore
What is an Observational Study?
A study which observes differences in the explanatory
variable and then assesses whether these differences are
related to differences in the response variable
A CAUSAL CLAIM CANNOT BE MADE
Famous NZ Observational Study is the Dunedin
Multidisciplinary Health and Development Study (DMHDS)
.A study of 1037 children who were born in Dunedin, in 197273. Where they now live shown below.
Dru Rose, Angela Hawkins and Rachel Passmore
Potential Problems with Observational
Studies
Confounding Variables
“Happy People Live
Longer”
• Possible Confounding variables –
Health, Life-style, Economic
status, Family………
• Why is this an Observational
Study? Cannot force people to
be happy !
• Problem – a causal claim has
Dru Rose, Angela Hawkins and Rachel Passmore
Potential Problems with Observational
Studies
Extending Results inappropriately
“WOMEN HIT HARDER BY HANGOVERS”
PROBLEM
The study observed only
University so results cannot be
extended to cover ALL WOMEN
Dru Rose, Angela Hawkins and Rachel Passmore
Potential Problems with Observational
Studies
Using the past as a source of data
“LEFT HANDED PEOPLE DIE
YOUNGER”
In early part of twentieth
century children forced to
write with right hands, hence
greater numbers of righthanders.
PROBLEM Confounding
variables can change over
time.
Recalled information often
unreliable.
Dru Rose, Angela Hawkins and Rachel Passmore
COMPARISON OF EXPERIMENTS AND
OBSERVATIONAL STUDIES
Experiments
Observational Studies
Explanatory Variable specified
Explanatory variable specified
Response variable specified
Response variable specified
Participants randomly allocated to treatments
Participants NOT randomly allocated to treatments,
sometimes because it is unethical to do so
Treatment is the only factor that is varied.
Changes in response variable recorded
Changes in response variable recorded
Good experiments include random allocation to
treatments, control groups, placebos and use
blinding
Good observational studies acknowledge & account for
all potential problems
Dru Rose, Angela Hawkins and Rachel Passmore
Writing frame for critically evaluating a
report
Read the media report and summarise what it is about in 3 sentences or bullet points.
MEDIA REPORT EVALUATION
Read the media report again, asking appropriate “worry questions” as you go.
Source
Observational Unit
Explanatory variable
Response variable
Treatment(s)
Study Type
Original Source
Design of study - describe
methodology of observational study
Method
Target Group
Who was sampled?
Selection method
OBSERVATIONAL STUDY - potential
problems
Sample size
Margin of error
Evaluate critical components
Key Findings
What is missing?
Critical Evaluation:
Discuss 2 good aspects of this report
CONCLUSION -
good or bad study? Any areas of
improvement?
Discuss 2 concern
Dru Rose, Angela Hawkins and Rachel Passmore
EXAMPLE OF EXAM STYLE QUESTION
QUESTION TWO
Extract from Newspaper article
College graduates feel better emotionally and
physically than their high-school drop-out
counterparts, a study of over 400,000 US adults
reveals. Participants were asked a series of
questions including level of education and on how
many of the past 30 days they felt physically and
emotionally healthy.
•Define the explanatory and response variables
•Explain whether any of the potential problems that
observational studies can suffer from might apply to this
study.
•Evaluation steps 5, 6 & 7 are concerned with
measurements taken and who was studied. Discuss
these steps in relation to this study
Dru Rose, Angela Hawkins and Rachel Passmore
POSSIBLE RESPONSES TO QUESTION
1. Explanatory Variable is level of education, Response variable is
Emotional & Physical Health
2. Recall of information over last 30 days may be unreliable, generalising
to ‘ good for your health’ but study only in US, confounding variables –
many other factors contribute to Emotional & Physical Health – social
network, economic status, family, illness etc
3. No information provided about who participants were or how they
were selected other than that they were from USA. Unclear exactly
what questions were asked to determine how healthy students were so
not sure if they provided a good indicator of health or not.
Responses to these questions might vary depending on when the
question was asked & how the student was feeling at that time. They
might also vary depending on who was asking the questions or how
responses were collected.
Dru Rose, Angela Hawkins and Rachel Passmore
```