Presentation on theme: "Updatedl 14 Nov: Turned link graph into vertical."— Presentation transcript:
1Conduct an experiment to investigate a situation using statistical methods Level 2 AS91265 3 credits Updatedl 14 Nov: Turned link graph into vertical.Updated: 7 Aug: Small tweaks to Letter-Spotting and Paired graphsCreated: 28 Jul-6 Aug 2012, MCS, Otumoetai College. Based on multiple sources, including Anna Martin’s notes “Auckland Mathematics Association 2011 Experiments at Level 7 workshop material” (Avondale College, and direction from nzstatsedn wiki – Experiments Level 7MCS: 14 Nov 2012
2This photo of a young Afghan girl Sharbat Gula has become an iconic image over the years since it featured on the cover of National Geographic magazine in The photograph was taken at the Nasir Bagh refugee camp in 1984 by photographer Steve McCurry. The girl was one of the students in an informal school within the refugee camp; McCurry, rarely given the opportunity to photograph Afghan women, seized the opportunity and captured her image. She was approximately 12 years old at the time.
3Honesty 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.Handout: Newcastle Honest experiment BBC Article.docx. Read this first
4Honesty 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.
5Controlled 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 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.This is described as the “first” controlled experiment. Why did Lind limit subjects to men who were similar?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.
6Controlled 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.Lind was following a line of though that acid could help cure scurvy, hence the reliance on acidic treatments. How was Lind controlling for other variables? How could he tell is the treatment worked?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.
8Wall-Sit ExperimentWe 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.We will use “A or B treatment allocation page.docx”, printed and cut out to allocated to groupsHow 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?
9Wall-Sit ExperimentA “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?Eliminate other sources of variation by ensuring that everyone does the experiment the same way. Students may opt out of this experiment if they are unable to do it. You may wish to use a stopwatch and call out the times. Students should remember their ‘time’, as we will record this.The treatment group should take some time to stretch, especially their quads.Both groups should start the exercise at the same time.
10Wall-Sit Experiment How can we quantify the effect of the treatment? How could we display the data?Normative data for the Wall Squat TestThe following table is the USA norms for 16 to 19 year olds.GenderExcellentAbove AverageAverageBelow AveragePoorMale>102 ssss<30 sFemale>60 ssss<20 sQuantify the effect by comparing averages (medians). Display with comparative dot plots
11What 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.
12What 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.At this level, we will imply causation in the treatment’s effect on the response variable, as we have attempted to control for other variables.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.
14Geography 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 possibleGroup 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.Random allocation to group A, group B using the paper random allocation paper cut-outs. “A or B treatment allocation page.docx” Students will write on the back of that piece of paper and these will be collected in.
15Group 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.____________Put heads down on desk and cover your paper when done.
16Now, group A close their eyes and group B open their eyes. Remember: No talking!Wait until all of group A have their heads down, and get group B to look up.
17Group 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.____________Cover your paper when done
18“Geography” Experiment We will collect in the pieces of paper and reveal what question the other group got.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?Collect in the data. Students answer these questions while the data is recorded. Use as practice for comparing two dot plots. Make a conclusion based on what you see.Answer: There are 55 countries in Africa (depending on the definition of country, and this includes South Sudan) – July 2012We are testing to see if the suggested “more than” number affects people’s guess. Explanatory: the suggested number Response: the guessed number.
19Independent 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”.We make a conclusion about this experiment, not an inference about a population from a sample.
20Pencil-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.Another Control/Treatment test. Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)Use: Pencil-in-mouth test sheets.docx
21Slide 1Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)
22Slide 2Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)
23Slide 3Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)
24Slide 3Slide 4Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)
25Slide 3Slide 5Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)
26Slide 3Slide 6Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)
27Slide 3Slide 7Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)
28Slide 3Slide 8Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)
29Slide 3Slide 9Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)
30Slide 3Slide 10Rate funny pictures on scale of 1 to 10 (1= not at all funny, 10=hilarious)
31Pencil-in-mouth Experiment Total up the score from each slide and write at the bottom.What do you think this experiment is testing?What is the treatment? What is the response variable?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 statisticsCompare what you seeWrite a conclusion:“The data suggests that…”Treatment: Pen in mouth. Response variable: perceived funniness
32Clinical TrialsExamples of research candidates wanted ads. Clinical trials need to be rigorous, so extra steps to reduce variation (see next slide)
33Blind 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.– discuss placebo effect“Placebo” – from the Latin “I shall please”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.
34PlacebosIn 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.– discuss placebo effect“Placebo” – from the Latin “I shall please”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”.
36Cola Wars of the mid-70’s onward Cola Wars of the mid-70’s onward. Pepsi did blind taste tests across the US and showed that more people actually preferred Pepsi. Coke hit back with similar test… but the result showed the same. So they reformulated Coke into “new Coke” which tasted more like Pepsi. Massive backlash resulted, and reverted a few years later to Classic Coke
37Pepsi ChallengeThe 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.
38“New” CokeSome 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.
39Letter-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.Paired-data comparison test.Use: Letter spotting sheet.docx
40Letter-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.Issue paper to everyone with above paragraph on it. One attempt at counting F’s (single read-through), record count and turn paper over.Then explain to class that the letter “F” sometimes sounds like a “V”. Count again.A Paired comparison (Before/After) to see if the knowledge of “F sounds like V” makes a difference.How many Fs did you count?__________Turn your paper over when done
41Letter-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.Issue paper to everyone with above paragraph on it. One attempt at counting F’s (single read-through), record count and turn paper over.Then explain to class that the letter “F” sometimes sounds like a “V”. Count again.A Paired comparison (Before/After) to see if the knowledge of “F sounds like V” makes a difference.How many Fs did you count on your second attempt?__________Turn your paper over when done
42Paired 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?)Note that we calculate the average of the differences (paired), not the difference of the averages (independent)
43Link Graphs Represent the finding with a link graph on the same scale. Attempt 1Attempt 2Baskets810911131214151719Represent the finding with a link graph on the same scale.Make sure the data is kept as a pair, joined with an arrow.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.Link graphs can be draw vertically or horizontally. The order of the attempts (top/bottom) doesn’t matter.
44Dot Plot of Differences A dot plot of differences:Attempt 1Attempt 2BasketsDifference810291113312-11141517194Difference = treatment – baseline.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.median2mean1.64modeLQUQ3IQR
45Letter-Spotting Experiment Counting the total number of F’s before and after the treatmentWhat do you think this experiment is testing?What is the treatment? What is the response variable?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 statisticsCompare what you seeWrite a conclusion:“The data suggests that…”Treatment: Knowledge of “F sounds like V” . Response variable: Number of F’s foundNOTE: There are 47 F’s in the paragraph
46Alphabet 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.Paired-data comparison test.Use: Alphabet Practice Experiment.docx
47Alphabet Practice Experiment Time how long it takes to find the letters A to Z in order.A Paired comparison (Before/After) to see if practice makes a difference.Issue paper to everyone with the above printed on it and 6 spaces to write times. Use stopwatch on projector or have their neighbour time them.
48Alphabet Practice Experiment Recording times to find the letters of the alphabet in orderWhat do you think this experiment is testing?What is the treatment? What is the response variable?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 statisticsCompare what you seeWrite a conclusion:“The data suggests that…”Treatment: Practice. Response variable: Time to find whole alphabet in order
49PPDAC CycleThis standard is concerned with all parts of the cycle
50ProblemUnderstand 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.
51P 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 measuredDescribe how the data will be collected and recordedIdentify 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
52DataWhen 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.
53Problem and Planning Practice Pick an investigation:Does eating chocolate increase your heart rate?Does the surface that you run on affect how fast you can run?Define the Problem that you will investigate :Define the exact problemWhat variables will be investigated?What is your hypothesis? (ie what do you think that you will find?)Write up a Plan for doing the experimentWhat do you know about the context?Define the exact explanatory variableDefine the response variable and how it will be measuredHow 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
54A 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)
55ConclusionDescriptive: "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?)
56C 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.
57Making 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.Repetition?
58Making a Conclusion Can we refute a claim by experimentation? ConclusionsHow could we refute a claim based on an experiment? – Seeing no (real) difference between control and treatment groups
59Paired comparison (dependent) Type of experimentPaired comparison (dependent)Comparison of two (or more) independent groupsQuestionsDoes doing something improve something? What’s the difference or change if I do something? What’s the effect of doing this?Is this better than this? Does doing this give better results than doing this? Does it matter if I do this?ConditionsSame group, before and after, measuring change/difference/ improvement in variable [linking two measurements of one variable from the same unit/person]Two different groups, comparing one variable across two (or more) independent measures/conditions [treatments]What are you manipulating?The in between, the change in conditions e.g. the fact that you exercise, drinking caffeine, watching a scary movieThe treatment(s) - what each group gets done to them (or for a control group, nothing)Exploratory data analysisLink-graph, dot plots of differences.Derive variable for increase/difference.Conclusion about change/improvement.Comparative dot plots and box plots, with summary statistics. Can explore data further and look at variation – shape, distance.What not to do?Don’t: Separate the two measurements and compare (breaking the link)Don’t: Compare two groups that are not independentExamplesHeart 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?Adapted from Anna Martin’s 2011 AMA notes
60- it’s a joke about the Batman villain called “two-face”, who has half a scarred face and half normal. He flips a coin as part of his schtick.