L2 Experiments AS-91265

Report
Conduct an
experiment to
investigate a
situation using
statistical methods
Level 2 AS91265
3 credits
MCS: 14 Nov 2012
Honesty Box Experiment
An experiment was carried out in a staff common room at Newcastle
University in England. There was an ‘honesty box’ system in place to
pay for tea/coffee use. Each week, an A5 poster was put above the
box. One week it has a picture of flowers, the other it has a picture of a
pair of eyes.
Honesty Box Experiment
There were about 48 staff who
used the common room, and the
honesty box system has been in
use for years, so users had no
idea that an experiment was
taking place.
Each week the research team
recorded the total amount of
money collected and the volume
of milk consumed as this was
considered to be the best index
available of total drink
consumption.
The team then calculated the
ratio of money collected to the
volume of milk consumed in
each week. On average, people
paid 2.76 as much for their
drinks on the weeks when the
poster featured pictures of eyes.
Controlled experimentation on scurvy
In 1747, while serving on the HMS Salisbury, surgeon James Lind
carried out a controlled experiment to develop a cure for scurvy.
Scurvy was a problem for sailors. During Anson's circumnavigation of
the world in 1740, he lost 1400 men out of an original crew of 1900 most of them allegedly from having contracted scurvy.
Unknown at the time, scurvy is a disease resulting from a deficiency of vitamin C.
Symptoms are malaise and lethargy, spots on the skin, bleeding gums and mucous
membranes. A person with scurvy looks pale, feels depressed, and is partially
immobilized. As it advances, there can be open, pus-forming wounds, loss of teeth,
jaundice, fever and death.
Lind selected 12 men from the ship, all suffering from
scurvy. Lind limited his subjects to men who "were as
similar as I could have them", that is he provided strict
entry requirements to reduce extraneous variation.
He divided them into six pairs, giving each pair different
supplements to their basic diet for two weeks.
Controlled experimentation on scurvy
The treatments were all remedies that had been
proposed:
• A quart of cider every day.
• Twenty five drops of elixir vitriol
(sulphuric acid) three times a day upon an
empty stomach.
• One half-pint of seawater every day.
• A mixture of garlic, mustard, and
horseradish in a golf-ball sized lump.
• Two spoonful's of vinegar three times a day.
• Two oranges and one lemon every day.
The men who had been given citrus fruits recovered
dramatically within a week. One of them returned to duty
after 6 days and the other cared for the rest.
The others experienced some improvement, but nothing
was comparable to the citrus fruits, which were proved to
be substantially superior to the other treatments.
Wall-Sit Experiment
We are going to conduct an experiment to see:
Does stretching before a wall-sit increase the length of time
that it can be done for?
For this experiment, we are going to divide the class into two
groups – the control group and the treatment group.
Control group: does exercise without any prior stretching.
Treatment group: stretches before doing exercise.
• How can we ensure that there is no bias in
the allocation of groups?
• Should we let people pick their own group?
• How will we know if the treatment works?
Wall-Sit Experiment
A “wall-sit” is done like this: stand with your back against a
wall and lower yourself until your thighs are parallel to the
floor. Lift your dominant foot about 5cm off the ground.
The experiment ends when the subject puts their foot back on
the ground.
• How do we eliminate other sources of variation?
• What kinds of variables could we measure in
this experiment?
The treatment group should take some time
to stretch, especially their quads.
Both groups should start the exercise at the
same time.
Wall-Sit Experiment
• How can we quantify the effect of the treatment?
• How could we display the data?
Normative data for the Wall Squat Test
The following table is the USA norms for 16 to 19 year olds.
Gender
Excellent
>102 s
Above
Average
102 - 76 s
Male
Female
>60 s
60 - 46 s
Average
75 - 58 s
Below
Average
57 - 30 s
Poor
<30 s
45 - 36 s
35 - 20 s
<20 s
What is an Experiment?
Experiment: A study in which a researcher attempts to understand the
effect that a variable (the explanatory variable) may have on some
phenomenon (the response) by controlling the conditions of the study.
The researcher controls the conditions by allocating individuals to
groups and allocating the value of the explanatory variable to be
received by each group. A value of the explanatory variable is called a
treatment.
In a well-designed experiment, the allocation of subjects to groups is
done using randomisation. Randomisation attempts to make the
characteristics of each group very similar so that if each group was
given the same treatment, the groups should respond in a similar way,
on average.
• Give an example for each word or phrase in red, using the wall-sit
experiment that we did.
What is an Experiment? Pt 2
Experiments usually have a control group, a group that receives no
treatment or receives an existing or established treatment. This allows
any differences in the response, on average, between the control group
and the treatment group(s) to be visible.
When the groups are similar in all ways apart from the treatment
received, then any observed differences in the response (if large
enough) among the groups, on average, is said to be caused by the
treatment.
Example: in an experiment to test
the effect of a fertilizer, the
experiment can be controlled by
assigning the treatments to
randomly selected plots of land.
This mitigates the effect of
variations in soil composition on
the growth of the plants and allows
us to see if the fertilizer works.
Geography Experiment
• For this experiment, we need two groups, A and B.
• You will be randomly allocated to a group.
• You will need a pen/pencil to write an answer on the
back of the piece of paper you get with your group on it.
• When the experiment begins, everyone must remain silent.
• Group A will look at the projector and answer the two questions
on it as quickly as possible
• Group B will have their eyes closed and head down on the desk.
• When instructed, Group A will close their eyes and Group B will
answer the questions on the projector.
Group A
Do you think there are more than 15 countries in
Africa?
Write: “Yes” or “No”
Now write down how many countries you think
there are in Africa.
____________
Now, group A close their eyes and
group B open their eyes.
Remember: No talking!
Group B
Do you think there are more than 50 countries in
Africa?
Write: “Yes” or “No”
Now write down how many countries you think
there are in Africa.
____________
“Geography” Experiment
We will collect in the pieces of paper and reveal what question
the other group got.
1.
2.
3.
4.
5.
6.
What do you think this experiment is testing?
What is the explanatory variable?
What is the response variable?
How will we compare the results?
What do you think the outcome will be? (“hypothesis”)
Why was it important that people were randomly allocated
to groups? That is, how are we trying to reduce variation
from variables other than the one that we are testing?
Independent Group Experiments
The Honesty-Box, Scurvy, Wall-sit and Geography experiments
are comparisons between independent groups.
These comparisons can be treatment vs control (stretching vs
not) or treatment 1 vs treatment 2 (seawater vs lemon juice).
Analysis:
• Comparative dot plots/box plots on the same scale.
• Generate summary statistics (medians, means, IQR)
• Compare what you see (differences, clusters, variation)
Conclusion:
• Descriptive, "the data suggests that using (the treatment)
caused (the students to jump further) on average".
• Can calculate the difference of averages (medians, means)
for some quantification (how much further?).
• NOT making an “inference about the population”.
Pencil-in-mouth Experiment
• For this experiment, we need two
groups, treatment and control.
• You will be randomly allocated to a
group.
• The treatment group will conduct
the experiment with a pen or pencil
between their teeth.
• You will need another pen to write
on your recording sheet.
• 10 slides will appear.
• Rate each slide for its funniness on
a scale from 1-10 where 1=not funny
and 10=hilarious.
Slide 1
Slide 2
Slide 3
Slide 4
Slide 3
Slide 5
Slide 3
Slide63
Slide
Slide
Slide73
Slide
Slide83
Slide
Slide93
Slide
Slide103
Pencil-in-mouth Experiment
Total up the score from each slide and write at the bottom.
1. What do you think this experiment is testing?
2. What is the treatment? What is the response variable?
3. What do you think the outcome will be? (Hypothesis)
The scores will be written on the board.
Analyse the results by:
• Constructing comparative dot
plots/box plots on the same scale.
• Generate summary statistics
• Compare what you see
Write a conclusion:
• “The data suggests that…”
Clinical Trials
Blind and Double-Blind Experiments
In blind experiments, at least some information is
withheld from participants in the experiments (but
not the experimenter).
For example, the subject might not know what the
experiment is testing for. Or there might be a
placebo given to the control group, and
participants don’t know which group they are in.
In double-blind experiments, both participants and
experimenters have limited information while the
experiment is being carried out.
Double-blind experiments common in clinical trials
of medical treatments, to verify that the supposed
effects of the treatment are produced only by the
treatment itself.
Placebos
In double-blind experiments using placebos, The placebo is the "first"
blind, and controls for the patient expectations that come with taking a
pill, which can have an effect on patient outcomes.
The "second" blind, of the experimenter, controls for the effects on
patient expectations due to unintentional differences in the
experimenter’s behaviour. Since the experimenter does not know
which patients are in which group, they cannot unconsciously
influence the patients.
After the experiment is over, they
then "unblind" themselves and
analyse the results.
Sometimes experiments reveal
unexpected improvements in the
placebo group as well as the
treatment group.
This is called “the placebo effect”.
Pepsi Challenge
The challenge took the form of a blind
taste test. At malls, shopping centres and
other public locations, a Pepsi
representative sets up a table with two
blank cups: one containing Pepsi and one
with Coca-Cola. Shoppers are encouraged
to taste both and then select which drink
they prefer. Then the representative
reveals the two bottles so the taster can
see whether they preferred Coke or Pepsi.
The results of the test suggested that
Pepsi was preferred by more Americans.
When cola taste samples include labels of Coke or
Pepsi the preference for Pepsi is reversed. Much of
the difference in preference for Coke is accounted
for by the label and not taste.
“New” Coke
Some suggest that Pepsi's success over Coca-Cola in the "Pepsi
Challenge" is a result of the flawed nature of the "sip test" method.
Malcolm Gladwell’s research shows that tasters will generally prefer
the sweeter of two beverages based on a single sip, even if they
prefer a less sweet beverage over the course of an entire can.
Just because a taster prefers a single sip of a sweeter beverage,
doesn't mean he or she would prefer to have an entire case of it at
home.
Regardless, the taste test results had Coca-Cola worried,
so in 1985 it reformulated the Coke recipe and replaced it
with “New Coke”. The taste tests agreed it was better, but
the public backlash was harsh.
Within three months, “Coca-Cola Classic” was back on
the shelves, and sales went higher than ever.
New Coke was eventually branded Coke II
from 1990 until it disappeared from shelves in 2002.
Letter-Spotting Experiment
• For this experiment, we use one
group, who will perform two
experiments with a treatment in the
middle.
• You will be issued with a piece of
paper with the task on it. Do not
turn it over until instructed.
• You will need a pen to write on your
recording sheet.
Letter-Spotting Experiment
Read this paragraph ONCE. As you do so, count how
many F’s you see.
The necessity of training farm hands for first class farms in the proficient
handling of farm livestock is foremost in the minds of effective farm
owners. Since the forefathers of the farm owners trained the farm hands
for first class farms in the proficient handling of farm livestock, the farm
owners feel they should carry on with the former family tradition of training
farmhands of first class farms in the effective handling of farm livestock,
however futile, because of their belief that it forms the basis of effective
farm management efforts.
How many Fs did you count?
__________
Turn your paper over when done
Letter-Spotting Experiment
Sometimes an ‘F’ sounds like a ‘V’. Read this paragraph
again ONCE. As you do so, count how many F’s you see.
The necessity of training farm hands for first class farms in the proficient
handling of farm livestock is foremost in the minds of effective farm
owners. Since the forefathers of the farm owners trained the farm hands
for first class farms in the proficient handling of farm livestock, the farm
owners feel they should carry on with the former family tradition of training
farmhands of first class farms in the effective handling of farm livestock,
however futile, because of their belief that it forms the basis of effective
farm management efforts.
How many Fs did you count on your second attempt?
__________
Turn your paper over when done
Paired Comparison Experiments
Unlike earlier experiments, the Letter-spotting experiment is a
comparison between paired data.
These comparisons are before (baseline) vs after treatment or
treatment 1 vs treatment 2. However, the data must be kept paired,
as it is not independent.
Analysis:
• A link graph on the same scale.
• Followed by a dot plot of differences.
• Compare what you see (direction of arrows in link graph,
clusters, variation, averages in dot plot of differences)
Conclusion:
• Descriptive, "the data suggests that (running for 5 minutes)
caused (heart rates to increase) on average".
• Can calculate the average of the differences (medians, means)
for some quantification (how much did they increase by?)
Link Graphs
• Represent the finding with a
link graph on the same scale.
• Make sure the data is kept as a pair,
joined with an arrow.
Attempt Attempt
1
2
Baskets Baskets
8
8
8
10
9
11
10
13
12
11
12
13
12
14
13
13
13
15
14
17
15
19
Compare what you see
(direction/gradient of arrows in link graph):
• The link graph shows that the vast majority of
people improved on their second attempt. We see
this as only one person scored fewer baskets on
the second attempt (arrow pointing down) and
only two people made no improvement (arrow
pointing straight across). The remaining 8 people
all improved (arrow pointing up).
• This suggests to us that the treatment (advice on
throwing a basketball) appears to improve the
number of baskets obtained out of 20 attempts.
Dot Plot of Differences
A dot plot of differences:
Difference in Basketball shot attempts
-1
0
1
2
3
Difference
Dot Plot
4
5
Attempt 1 Attempt 2
Baskets Baskets Difference
8
8
0
8
10
2
9
11
2
10
13
3
12
11
-1
12
13
1
12
14
2
13
13
0
13
15
2
14
17
3
15
19
4
Compare what you see (clusters, variation, averages):
• The differences are reasonably normally distributed,
centred around 2, which is the median of the
differences. The mean is 1.64, showing a slight skew
to the left.
• There are no significant clusters, but the mode is at a
difference of 2 baskets.
• The interquartile range shows the middle 50% of data
is between 0 and 3 extra baskets.
• This suggests to us that the treatment causes a
typical improvement of 2 extra successful baskets.
median
2
mean 1.64
mode
2
LQ
0
UQ
3
IQR
3
Letter-Spotting Experiment
Counting the total number of F’s before and after the treatment
1. What do you think this experiment is testing?
2. What is the treatment? What is the response variable?
3. What do you think the outcome will be? (Hypothesis)
The scores will be written on the board.
Analyse the results by:
• Constructing a link graph.
• Constructing a dot plot of differences.
• Generate summary statistics
• Compare what you see
Write a conclusion:
• “The data suggests that…”
Alphabet Practice Experiment
• For this experiment, we use one
group, who will perform an
experiment six times.
• You will be issued with a piece
of paper with the task on it. Do
not turn it over until instructed.
• You will need a pen to write on
your recording sheet.
• Each attempt needs to be timed,
and the time recorded.
Alphabet Practice Experiment
Time how long it takes to find the letters A to Z in order.
Alphabet Practice Experiment
Recording times to find the letters of the alphabet in order
1. What do you think this experiment is testing?
2. What is the treatment? What is the response variable?
3. What do you think the outcome will be? (Hypothesis)
The times for the first and last attempt will
be written on the board.
Analyse the results by:
• Constructing a link graph.
• Constructing a dot plot of differences.
• Generate summary statistics
• Compare what you see
Write a conclusion:
• “The data suggests that…”
PPDAC Cycle
roblem
• Understand and define the situation that
we are trying to investigate.
• Identify what variables will be investigated.
• Make sure you need an experiment to
investigate the relationship between them.
• What you would need to look for in the
data?
• Make a Hypothesis – predict the results of
the experiment.
lan
• Link to relevant knowledge about the context
• Explain how/why the explanatory variable will be
changed (could be categorical, eg control/ treatment)
• Explain how/why the response variable will be
measured
• Describe how the data will be collected and recorded
• Identify factors that might affect the results of the
experiment. ie Any related variables and the possible
effects of these (sources of variation/bias)
• List and justify the steps to be taken to carry out the
experiment
ata
• When carrying out experiments, make notes
about the data collection and experimental
process. These notes will be useful in
reflection of the process in the report write up.
• What else could you investigate that would help you
understand the situation better?
Or consider why your experiment did not give you the
expected results.
• You should be combining ideas of how a well run experiment
gives you good data, allowing you to answer the
investigation question knowing that the effects you observe
can be attributed to the variable you manipulated.
• Make sure that you clean the data if necessary.
Problem and Planning Practice
Pick an investigation:
1. Does eating chocolate increase your heart rate?
2. Does the surface that you run on affect how fast you can run?
Define the Problem that you will investigate :
• Define the exact problem
• What variables will be investigated?
• What is your hypothesis? (ie what do you think that you will find?)
Write up a Plan for doing the experiment
• What do you know about the context?
• Define the exact explanatory variable
• Define the response variable and how it will be measured
• How will the data be recorded?
• What other factors/variables/bias might affect the results?
• Describe in full how the experiment will be conducted.
• Justify all of your decisions
nalysis
Paired Comparisons:
• A link graph on the same scale.
• Followed by a dot plot of differences.
• Compare what you see (direction of arrows in link graph,
clusters, variation, averages in dot plot of differences)
Independent Groups:
• Comparative dot plots/box plots on the same scale.
• Generate summary statistics (medians, means, IQR)
• Compare what you see (differences, clusters, variation)
onclusion
• Descriptive: "the data suggests that using
(the treatment) caused (the students to
jump further) on average".
Paired Comparison:
• Can calculate the average of the
differences (medians, means) for some
quantification (how much did they
increase by?)
Independent Groups:
• Can calculate the difference of averages
(medians, means) for some quantification
(how much further?)
onclusion
• What did you find out from the experiment?
• What does the data suggest?
• Consider what other questions could be
investigated that would give more insight
into the experimental situation.
• Discuss the impact of other sources of
variation on the experimental data and
explain how this can be seen in the data.
• Reflect on how well the experiment went
and aspects that could be changed so that
the question could be investigated better.
Making a Conclusion
• Repetition is key part of
good experiment design.
• Ideally, we would repeat the
experiment many times
(or with many subjects)
to confirm the findings.
Making a Conclusion
Can we refute a claim by experimentation?
Type of
experiment
Questions
Conditions
Paired comparison (dependent)
Does doing something improve
something? What’s the difference or
change if I do something? What’s the
effect of doing this?
Same group, before and after,
measuring change/difference/
improvement in variable
[linking two measurements of one
variable from the same unit/person]
The in between, the change in
What are you conditions e.g. the fact that you
manipulating? exercise, drinking caffeine, watching a
scary movie
Comparison of two (or more)
independent groups
Is this better than this? Does doing this
give better results than doing this?
Does it matter if I do this?
Two different groups, comparing one
variable across two (or more)
independent measures/conditions
[treatments]
The treatment(s) - what each group
gets done to them (or for a control
group, nothing)
Exploratory
data analysis
Comparative dot plots and box plots,
Link-graph, dot plots of differences.
with summary statistics.
Derive variable for increase/difference.
Can explore data further and look at
Conclusion about change/improvement.
variation – shape, distance.
What not to
do?
Don’t: Separate the two measurements Don’t: Compare two groups that are
and compare (breaking the link)
not independent
Examples
Heart rate before compared to after
exercise, how much do heart rates
increase by?
Can people who stretch perform an
exercise longer than people who don’t?

similar documents